HubSpot vs Twenty

Agent access

1. ai-native userPoint an agent at llms.txt or agent-oriented docs

   weight 2 · round to Twenty
   HubSpot~partialprobed6/10

   HubSpot's docs explicitly state that agents can 'Fetch the complete documentation index at: https://developers.hubspot.com/docs/llms.txt' and that any doc page can be turned into markdown by appending .md (corroborated by real evidence sources that are themselves .md URLs, e.g. hubspot-docs-47, hubspot-docs-57). However, independent probes found the top-level https://developers.hubspot.com/llms.txt and https://developers.hubspot.com/docs.md both return 404, raising doubt about the accessibility/discoverability of the claimed agent-oriented entry points at the paths most agents would guess. Missing for 10: a probe confirming the specific /docs/llms.txt path actually resolves, and clearer top-level discoverability (root llms.txt) rather than only nested doc-page .md conversion.

   • [claimed-docs] “Every endpoint reference page publishes its machine-readable spec inline ("## OpenAPI ... specs/2026-09/account-account-info-v2026-09.json G…”
   • [probe] “PROBE llms.txt: HTTP 404 at https://developers.hubspot.com/llms.txt”
   • [probe] “PROBE docs-md: HTTP 404 at https://developers.hubspot.com/docs.md”
   • [claimed-docs] “Retrieve detailed information about the account associated with the HubSpot portal.”

   proof ↗⚑ flag

   Twenty✓fullprobed9/10

   A direct probe confirms llms.txt is live at docs.twenty.com/llms.txt returning HTTP 200 with structured navigation content, and Twenty's docs are also organized in agent-friendly markdown (.md endpoints) referenced throughout the evidence pack. Missing for 10: no independent/community confirmation of an agent successfully consuming llms.txt in practice.

   • [probe] “PROBE llms.txt: HTTP 200 at https://docs.twenty.com/llms.txt # Twenty Documentation ## Getting Started ### Welcome - [Why Twenty](https:/…”

   proof ↗⚑ flag

2. ai-native userRun the product headlessly / in CI for automation

   weight 2 · round to HubSpot
   HubSpot~partialprobed6/10

   HubSpot exposes a CLI (`npm install -g @hubspot/cli && hs init`) and a full REST API (contacts, custom objects, webhooks) plus a machine-readable OpenAPI spec, all of which can be scripted and run non-interactively in CI/automation pipelines. However, there is no explicit documentation or example showing CI/CD integration, headless CLI usage in pipelines, or automated testing workflows tailored to AI-native/agentic CI use. missing for 10: explicit CI/CD pipeline examples, documented headless/non-interactive CLI flags, and independent evidence of running HubSpot automation in a build pipeline.

   • [claimed-docs] “Install the HubSpot CLI globally / Authorize your HubSpot account / Start building”
   • [claimed-docs] “npm install -g @hubspot/cli && hs init”
   • [claimed-docs] “the contacts endpoints allow you to create and manage contact records in your HubSpot account, as well as sync contact data between HubSpot …”
   • [claimed-docs] “the webhooks journal API enables third-party integrators to subscribe to real-time events, retrieve historical event data, and manage snapsh…”
   • [claimed-docs] “Every endpoint reference page publishes its machine-readable spec inline ("## OpenAPI ... specs/2026-09/account-account-info-v2026-09.json G…”
   • [probe] “official MCP server documented at https://developers.hubspot.com/ai-tools/mcp”

   proof ↗⚑ flag

   Twenty~partialclaimed4/10

   Twenty can be self-hosted via a single Docker Compose command and exposes full REST/GraphQL APIs (including a Metadata API) for programmatic, UI-less interaction, which supports scripted/automated deployment and data operations. However, there is no explicit documentation of a headless mode, CI pipeline integration, or automated testing workflow — the evidence only covers deployment and API access, not CI-oriented usage. Missing for 10: explicit CI/CD examples, headless execution mode, automated test-runner support, and independent confirmation of running unattended in pipelines.

   • [claimed-docs] “Run Twenty on your own infrastructure with a single Docker Compose command.”
   • [claimed-docs] “Install the latest stable version of Twenty with a single command: bash <(curl -sL https://raw.githubusercontent.com/twentyhq/twenty/main/pa…”
   • [claimed-docs] “This guide provides step-by-step instructions to install and configure the Twenty application using Docker Compose.”
   • [claimed-docs] “A developer-first API that adapts to your custom data model. Both GraphQL and REST endpoints, with auto-generated documentation per works…”
   • [claimed-docs] “when you add a custom object (say Invoice), it immediately gets REST and GraphQL endpoints identical to built-in objects like Company or Per…”
   • [claimed-docs] “Metadata API — /rest/metadata/ and /metadata/ Schema management: create/modify/delete objects, fields, and relations. This is how you progra…”

   proof ↗⚑ flag

3. ai-native userPlug MCP servers into this product so it can use their tools

   weight 3 · round drawn
   HubSpot—none0/10

   All MCP evidence describes HubSpot shipping its own MCP server so external AI tools/agents can call into HubSpot's CRM and developer platform (hubspot-docs-5, hubspot-docs-30, hubspot-docs-40, hubspot-docs-6, hubspot-docs-24, hubspot-probe-4) — i.e., HubSpot as the MCP server, not as an MCP client. The story asks the reverse: can a user plug external MCP servers into HubSpot so HubSpot's own AI features can use their tools. No evidence shows HubSpot's Breeze/AI agents consuming or configuring third-party MCP servers as tool sources.

   • [claimed-docs] “The HubSpot MCP server enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — conta…”
   • [claimed-docs] “The HubSpot MCP server enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data”
   • [claimed-docs] “enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — contacts, deals, engagements…”
   • [claimed-docs] “The Developer MCP server enables your agentic development tools to interact with the HubSpot Developer Platform via the CLI.”
   • [claimed-docs] “This server is CLI-based and enables your agentic development software to interact with the HubSpot Developer Platform, helping you to quick…”
   • [probe] “official MCP server documented at https://developers.hubspot.com/ai-tools/mcp”

   proof ↗⚑ flag

   Twenty—none0/10

   The only MCP-related evidence is a pricing table line item 'MCP server | Yes' (twenty-docs-32), which indicates Twenty exposes an MCP server for others to call its own tools — not that Twenty acts as an MCP client that can plug in external MCP servers so its AI agents can use their tools. No documentation describes configuring or connecting external MCP servers within Twenty's AI agent framework (twenty-docs-5, twenty-docs-18, twenty-docs-56 describe agents but only mention internal actions like enrichment, not external MCP tool integration).

   • [claimed-docs] “MCP server | Yes | Yes | Yes”
   • [claimed-docs] “AI agents that work autonomously within your CRM — answering questions, enriching records, and executing multi-step tasks within your permis…”
   • [claimed-docs] “Ask questions about your data in natural language. The AI chatbot can query your CRM records, summarize information, and help you find what …”
   • [claimed-docs] “AI agents go beyond chat — they can execute multi-step tasks autonomously: Enrich records with data from external sources”

   proof ↗⚑ flag

4. ai-native userConnect an agent via an official MCP server

   weight 3 · round to HubSpot
   HubSpot✓fullprobed8/10

   HubSpot documents an official MCP server enabling MCP-compatible AI tools/agents to get secure read/write access to CRM data, plus a separate Developer MCP server for agentic dev tools, confirmed by probe-4 finding the docs page live. Missing for 10: independent/hands-on third-party corroboration of the MCP server working in practice, and no llms.txt/openapi discovery worked at the root domain (probes 1-3 failed), though the endpoint itself is documented.

   • [claimed-docs] “The HubSpot MCP server enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — conta…”
   • [claimed-docs] “The Developer MCP server enables your agentic development tools to interact with the HubSpot Developer Platform via the CLI.”
   • [claimed-docs] “This server is CLI-based and enables your agentic development software to interact with the HubSpot Developer Platform, helping you to quick…”
   • [claimed-docs] “The HubSpot MCP server enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data”
   • [claimed-docs] “enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — contacts, deals, engagements…”
   • [probe] “official MCP server documented at https://developers.hubspot.com/ai-tools/mcp”

   proof ↗⚑ flag

   Twenty✓fullclaimed5/10

   Twenty's pricing page explicitly lists 'MCP server | Yes | Yes | Yes' across all plans, confirming an official MCP server offering that would let AI agents connect. However, this is the only evidence — there's no dedicated documentation page explaining setup, capabilities, or authentication for the MCP server, and no independent/hands-on corroboration of it working. Missing for 10: dedicated MCP server docs/setup guide, independent verification that it works, details on scope/permissions for MCP connections.

   • [claimed-docs] “MCP server | Yes | Yes | Yes”

   proof ↗⚑ flag

5. ai-native userUse an official CLI

   weight 2 · round to HubSpot
   HubSpot✓fullclaimed8/10

   HubSpot ships an official CLI (`npm install -g @hubspot/cli && hs init`), documented with install/auth/build steps, and it's a first-class tool used to scaffold and manage projects, including agentic development flows via the Developer MCP server which itself is CLI-based. Missing for 10: independent/hands-on developer corroboration of CLI usage and depth of CLI command coverage beyond install/init.

   • [claimed-docs] “Install the HubSpot CLI globally / Authorize your HubSpot account / Start building”
   • [claimed-docs] “npm install -g @hubspot/cli && hs init”
   • [claimed-docs] “This server is CLI-based and enables your agentic development software to interact with the HubSpot Developer Platform, helping you to quick…”
   • [claimed-docs] “The Developer MCP server enables your agentic development tools to interact with the HubSpot Developer Platform via the CLI.”

   proof ↗⚑ flag

   Twenty✓fullclaimed7/10

   Twenty ships an official CLI (`npx create-twenty-app`, `npx twenty app:publish`) documented in the GitHub README for scaffolding apps and defining objects/fields/views as code, satisfying the 'official CLI' axis. Missing for 10: independent/hands-on corroboration of the CLI's reliability, and documentation of broader CLI capabilities beyond app scaffolding/publishing (e.g., data or workflow management from the CLI).

   • [github] “Scaffold a new app with the Twenty CLI: npx create-twenty-app my-app ... Define objects, fields, and views as code”
   • [github] “npx twenty app:publish --private”
   • [github] “npx create-twenty-app my-app ... npx twenty app:publish --private”
   • [github] “Scaffold a new app with the Twenty CLI: npx create-twenty-app my-app”
   • [github] “Define objects, fields, and views as code”

   proof ↗⚑ flag

6. ai-native userDrive the product through a documented public API

   weight 3 · round drawn
   HubSpot✓fullprobed8/10

   HubSpot publishes extensive REST API documentation (CRM objects, contacts, custom objects, webhooks, account info) with concrete endpoint examples (e.g. POST /crm/v3/objects/contacts), a CLI, versioning/deprecation policies, and inline OpenAPI specs per endpoint, plus llms.txt/markdown-friendly docs for agent consumption. missing for 10: independent/hands-on developer corroboration that the public API works as documented in practice, and the root-level llms.txt/openapi probes returned 404 (though a nested docs.md/llms.txt path is referenced in supp evidence), leaving some ambiguity about universal machine-readability.

   • [claimed-docs] “To create one contact, make a `POST` request to `/crm/v3/objects/contacts`.”
   • [claimed-docs] “the contacts endpoints allow you to create and manage contact records in your HubSpot account, as well as sync contact data between HubSpot …”
   • [claimed-docs] “HubSpot custom objects allow organizations to represent and organize data based on unique business requirements. Create and update custom ob…”
   • [claimed-docs] “Retrieve detailed information about the account associated with the HubSpot portal.”
   • [claimed-docs] “Docs, "2026-09 API reference": "The 2026-09 APIs use a date-based versioning scheme (e.g., /crm/2026-09/)... This versioning model gives Hub…”
   • [claimed-docs] “Docs, "Sunsetted and deprecated APIs": a maintained table of sunsetted APIs with dates and migration targets, plus "Deprecated APIs... HubSp…”
   • [claimed-docs] “Every endpoint reference page publishes its machine-readable spec inline ("## OpenAPI ... specs/2026-09/account-account-info-v2026-09.json G…”
   • [probe] “PROBE llms.txt: HTTP 404 at https://developers.hubspot.com/llms.txt”
   • [probe] “PROBE openapi: all candidate paths 404 (https://developers.hubspot.com/openapi.json, https://developers.hubspot.com/swagger.json, https://de…”

   proof ↗⚑ flag

   Twenty✓fullprobed8/10

   Twenty documents a developer-first public API with both REST and GraphQL endpoints, auto-generated per-workspace docs, a Metadata API for programmatic schema changes, batching support, and scoped API keys — plus community confirmation that the REST API works hands-on. missing for 10: independent verification of documentation completeness/quality (probe found no discoverable OpenAPI spec at standard paths) and more third-party corroboration beyond one HN comment.

   • [claimed-docs] “A developer-first API that adapts to your custom data model. Both GraphQL and REST endpoints, with auto-generated documentation per works…”
   • [claimed-docs] “when you add a custom object (say Invoice), it immediately gets REST and GraphQL endpoints identical to built-in objects like Company or Per…”
   • [claimed-docs] “Metadata API — /rest/metadata/ and /metadata/ Schema management: create/modify/delete objects, fields, and relations. This is how you progra…”
   • [claimed-docs] “Keys can be scoped to a specific role under Settings → Members → Roles → Assignment tab to limit what they can access.”
   • [claimed-docs] “Both REST and GraphQL support batching up to 60 records per request — create, update, or delete. GraphQL also supports batch upsert”
   • [claimed-docs] “when you add a custom object (say `Invoice`), it immediately gets REST and GraphQL endpoints identical to built-in objects like `Company` or…”
   • [community] “Twenty was almost what we needed. It was extremely easy to set up, the customization is very promising, and the REST API worked ok. The main…”
   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.twenty.com/openapi.json, https://docs.twenty.com/swagger.json, https://docs.twenty.com/…”

   proof ↗⚑ flag

7. ai-native userIssue scoped/least-privilege API credentials for an agent

   weight 2 · round to Twenty
   HubSpot—none0/10

   The evidence pack documents HubSpot's MCP server giving broad 'secure read and write access' to CRM data and general account/API mechanics, but nowhere describes scoped, least-privilege API credentials (e.g., granular OAuth scopes or private-app permission scoping) that an AI-native user could issue specifically for an agent.

   • [claimed-docs] “The HubSpot MCP server enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — conta…”
   • [claimed-docs] “The HubSpot MCP server enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data”
   • [claimed-docs] “enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — contacts, deals, engagements…”
   • [claimed-docs] “You can create up to 10 test accounts to test apps and integrations without affecting any real HubSpot data.”

   proof ↗⚑ flag

   Twenty✓fullclaimed7/10

   Twenty's docs explicitly state API keys can be scoped to a specific role to limit access (twenty-docs-24), backed by granular RBAC at object/field/record level (twenty-docs-7, twenty-docs-39, twenty-docs-54) and an explicit permission model constraining AI agents to only what the user/role can view (twenty-docs-19, twenty-docs-5). This directly supports issuing least-privilege credentials for an agent. Missing for 10: independent/hands-on verification that role-scoped keys work as documented, and no explicit agent-specific key-issuance workflow beyond general role scoping.

   • [claimed-docs] “Keys can be scoped to a specific role under Settings → Members → Roles → Assignment tab to limit what they can access.”
   • [claimed-docs] “Role-based access control at every level — objects, fields, and individual records. Configure SSO with SAML or OIDC.”
   • [claimed-docs] “Role-based access control at every level — objects, fields, and individual records. Configure SSO with SAML or OIDC. Audit logs track who di…”
   • [claimed-docs] “Agents can only access objects and fields that the user (or role) has permission to view. Sensitive data stays protected even when AI is inv…”
   • [claimed-docs] “AI agents that work autonomously within your CRM — answering questions, enriching records, and executing multi-step tasks within your permis…”

   proof ↗⚑ flag

8. ai-native userBuild against official SDKs

   weight 2 · round to Twenty
   HubSpot—none0/10

   The evidence pack documents HubSpot's REST APIs, an OpenAPI-spec-per-endpoint, a CLI (@hubspot/cli), and MCP servers, but nowhere mentions official client SDKs/libraries (e.g., Node.js, Python, PHP, Ruby SDKs) that an AI-native developer would build against. Missing for 10: any documentation of official SDK packages, language coverage, or SDK release/versioning cadence.

   • [claimed-docs] “Install the HubSpot CLI globally / Authorize your HubSpot account / Start building”
   • [claimed-docs] “npm install -g @hubspot/cli && hs init”
   • [claimed-docs] “Every endpoint reference page publishes its machine-readable spec inline ("## OpenAPI ... specs/2026-09/account-account-info-v2026-09.json G…”
   • [claimed-docs] “Docs, "2026-09 API reference": "The 2026-09 APIs use a date-based versioning scheme (e.g., /crm/2026-09/)... This versioning model gives Hub…”

   proof ↗⚑ flag

   Twenty~partialprobed6/10

   Twenty offers a developer-first REST/GraphQL API with auto-generated docs, a Metadata API, an official CLI (`create-twenty-app`) for scaffolding apps as code, and an MCP server offering (listed on pricing), all of which support AI-native builders. However, there is no explicit 'official SDK' (e.g., published npm/pip client library) mentioned, and a probe found no discoverable OpenAPI spec, undercutting the auto-generated-docs claim. Missing for 10: named official SDK packages in specific languages, independent/hands-on confirmation of building against these APIs, and a working OpenAPI/schema endpoint.

   • [claimed-docs] “A developer-first API that adapts to your custom data model. Both GraphQL and REST endpoints, with auto-generated documentation per works…”
   • [claimed-docs] “when you add a custom object (say Invoice), it immediately gets REST and GraphQL endpoints identical to built-in objects like Company or Per…”
   • [claimed-docs] “Metadata API — /rest/metadata/ and /metadata/ Schema management: create/modify/delete objects, fields, and relations. This is how you progra…”
   • [github] “Scaffold a new app with the Twenty CLI: npx create-twenty-app my-app ... Define objects, fields, and views as code”
   • [github] “npx create-twenty-app my-app ... npx twenty app:publish --private”
   • [claimed-docs] “MCP server | Yes | Yes | Yes”
   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.twenty.com/openapi.json, https://docs.twenty.com/swagger.json, https://docs.twenty.com/…”

   proof ↗⚑ flag

9. ai-native userSubscribe to events via webhooks

   weight 2 · round to HubSpot
   HubSpot✓fullclaimed8/10

   HubSpot documents a dedicated webhooks journal API allowing third-party integrators to subscribe to real-time events, retrieve historical event data, and manage CRM object snapshots, directly matching the story. Missing for 10: independent/hands-on corroboration of webhook subscription reliability and no explicit mention of AI-agent-specific webhook consumption patterns.

   • [claimed-docs] “the webhooks journal API enables third-party integrators to subscribe to real-time events, retrieve historical event data, and manage snapsh…”

   proof ↗⚑ flag

   Twenty~partialclaimed5/10

   Twenty's workflow engine can trigger on record changes/schedules/incoming webhooks and chain an 'HTTP request' step (e.g., Slack alert on deal stage change), effectively enabling outbound webhook-style notifications for events — but there's no documented first-class 'webhook subscription' API where an external AI agent registers a URL to be notified of arbitrary CRM events. Missing for 10: explicit webhook subscription/registration endpoint or events API, documentation of supported event types/payloads for outbound webhooks, and independent confirmation this works reliably.

   • [claimed-docs] “Trigger workflows on record changes, schedules, manual actions, or incoming webhooks.”
   • [claimed-docs] “Workflows can chain any combination of: Record operations... Send email... HTTP request... Code... Branches... Iterator... AI Agent... Delay…”
   • [claimed-docs] “Send Slack alerts when a deal reaches a certain stage * Auto-enrich new contacts with data from external APIs”
   • [claimed-docs] “Workflows let you automate repetitive tasks and connect Twenty to external tools — without writing code (though you can if you want to).”

   proof ↗⚑ flag

Agentic features

1. ai-native userGet AI-generated insights and suggestions from my data inside the product

   weight 2 · round to Twenty
   HubSpot✓fullclaimed7/10

   HubSpot ships Breeze AI features embedded in the product: natural-language report generation ('describe the report you need in plain language, and HubSpot will build it from your actual data'), AI agents for prospecting/deal progression that surface insights and next-step suggestions from CRM data, and content-generation AI. These are native, in-product AI insight/suggestion features rather than external tooling. missing for 10: independent hands-on validation of insight quality/accuracy, and deeper detail on how proactive/contextual the insights are beyond marketing copy.

   • [claimed-docs] “describe the report you need in plain language, and HubSpot will build it from your actual data.”
   • [claimed-docs] “with Breeze Assistant, describe the report you need in plain language, and HubSpot will build it from your actual data.”
   • [claimed-docs] “Prospecting agent watches accounts for signals that they're ready to buy and reaches out automatically, while smart deal progression keeps d…”
   • [claimed-docs] “Prospecting agent watches accounts for signals that they're ready to buy and reaches out automatically, while smart deal progression keeps d…”
   • [claimed-docs] “Approve automatic CRM updates and follow-up drafts after every meeting”
   • [claimed-docs] “Build demand, win deals, and serve your customers with a team of AI agents for marketing, sales, and customer service. Built on your CRM dat…”
   • [claimed-docs] “HubSpot's free CRM unifies all your customer data on one platform, with AI that makes it easy to understand. And unlike a free trial, it's 1…”

   proof ↗⚑ flag

   Twenty✓fullclaimed8/10

   Twenty ships a natural-language AI chatbot that queries CRM records, summarizes information, and finds answers without building filters, plus AI agents that enrich records and execute multi-step tasks autonomously within permission boundaries. This directly delivers AI-generated insights/suggestions from the user's own data inside the product. Missing for 10: independent/hands-on validation of the AI chatbot and agent quality (only vendor docs), and no evidence of proactive/unprompted suggestions beyond query-driven and record-enrichment use cases.

   • [claimed-docs] “AI agents that work autonomously within your CRM — answering questions, enriching records, and executing multi-step tasks within your permis…”
   • [claimed-docs] “Ask questions about your data in natural language. The AI chatbot can query your CRM records, summarize information, and help you find what …”
   • [claimed-docs] “Agents can only access objects and fields that the user (or role) has permission to view. Sensitive data stays protected even when AI is inv…”
   • [claimed-docs] “AI agents go beyond chat — they can execute multi-step tasks autonomously: Enrich records with data from external sources”

   proof ↗⚑ flag

2. ai-native userSet up automations that run autonomously in the background

   weight 2 · round drawn
   HubSpot✓fullclaimed7/10

   HubSpot ships native Workflows automation that runs processes autonomously on CRM triggers, plus AI agents (e.g., the Prospecting agent) explicitly described as watching accounts for signals and reaching out automatically, and smart deal progression handling next steps without user action — all clearly background/autonomous behaviors. Webhooks and the CRM API further support event-driven automation an AI-native user could wire up. Missing for 10: independent/hands-on verification that these agents run reliably unattended over time, and deeper docs on scheduling/trigger configuration for developer-built autonomous flows.

   • [claimed-docs] “Create workflows in HubSpot to automate your processes.”
   • [claimed-docs] “Create workflows”
   • [claimed-docs] “Prospecting agent watches accounts for signals that they're ready to buy and reaches out automatically, while smart deal progression keeps d…”
   • [claimed-docs] “Prospecting agent watches accounts for signals that they're ready to buy and reaches out automatically, while smart deal progression keeps d…”
   • [claimed-docs] “Approve automatic CRM updates and follow-up drafts after every meeting”
   • [claimed-docs] “the webhooks journal API enables third-party integrators to subscribe to real-time events, retrieve historical event data, and manage snapsh…”
   • [claimed-docs] “Build demand, win deals, and serve your customers with a team of AI agents for marketing, sales, and customer service. Built on your CRM dat…”

   proof ↗⚑ flag

   Twenty✓fullclaimed7/10

   Twenty's workflow engine supports triggers on schedules, record changes, and webhooks, chaining actions including AI Agent steps, code execution, and delays—enabling automations that run autonomously in the background without user intervention. Docs further describe AI agents executing multi-step tasks autonomously (enrichment, chat) within a permission model. Missing for 10: independent/hands-on verification that scheduled/background workflows run reliably at scale, and detail on monitoring or retry/error handling for failed autonomous runs.

   • [claimed-docs] “Trigger workflows on record changes, schedules, manual actions, or incoming webhooks.”
   • [claimed-docs] “Workflows can chain any combination of: Record operations... Send email... HTTP request... Code... Branches... Iterator... AI Agent... Delay…”
   • [claimed-docs] “AI agents that work autonomously within your CRM — answering questions, enriching records, and executing multi-step tasks within your permis…”
   • [claimed-docs] “AI agents go beyond chat — they can execute multi-step tasks autonomously: Enrich records with data from external sources”
   • [claimed-docs] “Agents can only access objects and fields that the user (or role) has permission to view. Sensitive data stays protected even when AI is inv…”
   • [claimed-docs] “Workflows let you automate repetitive tasks and connect Twenty to external tools — without writing code (though you can if you want to).”

   proof ↗⚑ flag

3. ai-native userDelegate tasks to a built-in AI assistant inside the product

   weight 3 · round to Twenty
   HubSpot~partialclaimed7/10

   HubSpot documents built-in AI agents (Breeze Assistant, Prospecting agent, campaign-planning agent) that can autonomously watch accounts, draft follow-ups, build reports from plain-language requests, and generate content — a clear built-in delegate-to-AI-assistant capability (hubspot-docs-14, 22, 23, 35, 50, 56). However, all evidence is vendor-authored marketing/docs pages with no independent or hands-on corroboration of these agents actually performing delegated tasks reliably. Missing for 10: independent/hands-on verification of the AI assistant executing delegated tasks, and detail on scope/limits of autonomy.

   • [claimed-docs] “Prospecting agent watches accounts for signals that they're ready to buy and reaches out automatically, while smart deal progression keeps d…”
   • [claimed-docs] “Prospecting agent watches accounts for signals that they're ready to buy and reaches out automatically, while smart deal progression keeps d…”
   • [claimed-docs] “Approve automatic CRM updates and follow-up drafts after every meeting”
   • [claimed-docs] “describe the report you need in plain language, and HubSpot will build it from your actual data.”
   • [claimed-docs] “with Breeze Assistant, describe the report you need in plain language, and HubSpot will build it from your actual data.”
   • [claimed-docs] “Build demand, win deals, and serve your customers with a team of AI agents for marketing, sales, and customer service. Built on your CRM dat…”
   • [claimed-docs] “Create blog posts, social content, and landing pages in your brand voice”

   proof ↗⚑ flag

   Twenty✓fullclaimed8/10

   Twenty documents built-in AI agents that operate autonomously within the CRM—answering questions, enriching records, and executing multi-step tasks—scoped by the user's permission model, plus a natural-language AI chatbot for querying data, both delegated to within the product itself. Missing for 10: independent/hands-on verification of the AI assistant's real-world reliability and no community evidence corroborating this specific AI feature (community feedback predates the AI feature and is unrelated).

   • [claimed-docs] “AI agents that work autonomously within your CRM — answering questions, enriching records, and executing multi-step tasks within your permis…”
   • [claimed-docs] “Ask questions about your data in natural language. The AI chatbot can query your CRM records, summarize information, and help you find what …”
   • [claimed-docs] “Agents can only access objects and fields that the user (or role) has permission to view. Sensitive data stays protected even when AI is inv…”
   • [claimed-docs] “AI agents go beyond chat — they can execute multi-step tasks autonomously: Enrich records with data from external sources”
   • [claimed-docs] “Workflows can chain any combination of: Record operations... Send email... HTTP request... Code... Branches... Iterator... AI Agent... Delay…”

   proof ↗⚑ flag

4. ai-native userOperate the product with natural-language commands

   weight 2 · round to Twenty
   HubSpot~partialclaimed6/10

   HubSpot's Breeze AI features let users describe reports in plain language and plan campaigns from a single goal, and its MCP server exposes CRM data to natural-language-driven AI agents — this is genuine natural-language operation of parts of the product. However, coverage is limited to specific modules (reporting, campaigns, prospecting) rather than the whole platform, and there's no independent/hands-on evidence corroborating how well these NL commands work in practice. Missing for 10: broad NL control across core CRM workflows (contacts, deals, workflows) beyond reporting/campaigns, and independent verification of NL command reliability.

   • [claimed-docs] “describe the report you need in plain language, and HubSpot will build it from your actual data.”
   • [claimed-docs] “with Breeze Assistant, describe the report you need in plain language, and HubSpot will build it from your actual data.”
   • [claimed-docs] “Plan a full campaign from a single goal, in one session”
   • [claimed-docs] “Build demand, win deals, and serve your customers with a team of AI agents for marketing, sales, and customer service. Built on your CRM dat…”
   • [claimed-docs] “The HubSpot MCP server enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — conta…”
   • [claimed-docs] “enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — contacts, deals, engagements…”

   proof ↗⚑ flag

   Twenty✓fullclaimed7/10

   Twenty's docs describe a built-in AI chatbot that lets users 'ask questions about your data in natural language' to query records, summarize info, and avoid building complex filters, plus autonomous AI agents that execute multi-step tasks within the CRM under the user's permission model. This directly satisfies the natural-language operation story for an AI-native user. Missing for 10: independent/hands-on verification of the chatbot's reliability and no community reports confirming it works as documented in practice.

   • [claimed-docs] “Ask questions about your data in natural language. The AI chatbot can query your CRM records, summarize information, and help you find what …”
   • [claimed-docs] “AI agents that work autonomously within your CRM — answering questions, enriching records, and executing multi-step tasks within your permis…”
   • [claimed-docs] “AI agents go beyond chat — they can execute multi-step tasks autonomously: Enrich records with data from external sources”
   • [claimed-docs] “Agents can only access objects and fields that the user (or role) has permission to view. Sensitive data stays protected even when AI is inv…”

   proof ↗⚑ flag

Api quality

1. ai-native userExplore an interactive API reference with runnable examples

   weight 2 · round to HubSpot
   HubSpot~partialprobed4/10

   HubSpot publishes machine-readable API reference pages (OpenAPI specs, curl/code examples like the contacts POST endpoint, and agent-readable .md/llms.txt paths per hubspot-supp-openapi-spec, hubspot-docs-31, hubspot-docs-57), showing a structured API reference exists. However, there's no evidence of an interactive 'try it' console or runnable execution within the docs, and probes for llms.txt/openapi.json at expected root paths returned 404s, suggesting the interactive/agent-discoverable surface is inconsistent or non-standard. Missing for 10: explicit interactive 'run this call' sandbox UI, confirmation that root-level llms.txt/openapi.json resolve, and independent corroboration of hands-on interactivity.

   • [claimed-docs] “Every endpoint reference page publishes its machine-readable spec inline ("## OpenAPI ... specs/2026-09/account-account-info-v2026-09.json G…”
   • [claimed-docs] “Retrieve detailed information about the account associated with the HubSpot portal.”
   • [claimed-docs] “To create one contact, make a `POST` request to `/crm/v3/objects/contacts`.”
   • [probe] “PROBE llms.txt: HTTP 404 at https://developers.hubspot.com/llms.txt”
   • [probe] “PROBE openapi: all candidate paths 404 (https://developers.hubspot.com/openapi.json, https://developers.hubspot.com/swagger.json, https://de…”

   proof ↗⚑ flag

   Twenty~partialprobed3/10

   Docs mention a 'developer-first API' with 'auto-generated documentation per workspace' for both GraphQL and REST endpoints (twenty-docs-8), implying some interactive per-workspace API reference, but there is no explicit description of runnable/try-it-out examples, and a probe for a standard OpenAPI/Swagger endpoint at the docs domain returned 404s (inconclusive since it wasn't the per-workspace endpoint). Missing for 10: explicit confirmation of an interactive 'try it out' console, screenshots or docs describing runnable code examples, and independent corroboration that the auto-generated docs are interactive rather than static.

   • [claimed-docs] “A developer-first API that adapts to your custom data model. Both GraphQL and REST endpoints, with auto-generated documentation per works…”
   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.twenty.com/openapi.json, https://docs.twenty.com/swagger.json, https://docs.twenty.com/…”

   proof ↗⚑ flag

2. ai-native userDownload a machine-readable API spec (OpenAPI or equivalent)

   weight 2 · round to HubSpot
   HubSpot~partialprobed6/10

   First-party docs show each API reference page embeds a machine-readable OpenAPI 3.0.1 spec (e.g., specs/2026-09/account-account-info-v2026-09.json) and an agent-readable index at /docs/llms.txt, confirming a real OpenAPI-based spec exists per endpoint. However, independent probes for a single top-level downloadable spec (openapi.json, swagger.json, llms.txt at root) all returned 404, suggesting there is no one-shot full-API spec download — only per-endpoint fragments discoverable via docs navigation. Missing for 10: a consolidated single-file OpenAPI document confirmed downloadable, and independent/community verification of successfully fetching it.

   • [claimed-docs] “Every endpoint reference page publishes its machine-readable spec inline ("## OpenAPI ... specs/2026-09/account-account-info-v2026-09.json G…”
   • [probe] “PROBE llms.txt: HTTP 404 at https://developers.hubspot.com/llms.txt”
   • [probe] “PROBE docs-md: HTTP 404 at https://developers.hubspot.com/docs.md”
   • [probe] “PROBE openapi: all candidate paths 404 (https://developers.hubspot.com/openapi.json, https://developers.hubspot.com/swagger.json, https://de…”

   proof ↗⚑ flag

   Twenty~partialprobed5/10

   Docs state the API is REST + GraphQL with 'auto-generated documentation per workspace' (twenty-docs-8), implying some machine-readable spec exists, and the Metadata API can describe the schema programmatically (twenty-docs-23/43). However, no evidence confirms a concrete OpenAPI/Swagger file is downloadable, and a direct probe for standard OpenAPI paths on the docs site returned 404s, leaving the claim unconfirmed. Missing for 10: a documented/downloadable OpenAPI or GraphQL SDL export endpoint, and hands-on confirmation that the 'auto-generated documentation per workspace' is actually machine-readable (e.g., JSON schema) rather than just human-readable docs.

   • [claimed-docs] “A developer-first API that adapts to your custom data model. Both GraphQL and REST endpoints, with auto-generated documentation per works…”
   • [claimed-docs] “Metadata API — /rest/metadata/ and /metadata/ Schema management: create/modify/delete objects, fields, and relations. This is how you progra…”
   • [claimed-docs] “Metadata API — Schema management: create/modify/delete objects, fields, and relations. This is how you programmatically change your data mod…”
   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.twenty.com/openapi.json, https://docs.twenty.com/swagger.json, https://docs.twenty.com/…”

   proof ↗⚑ flag

3. ai-native userTest against a sandbox environment without touching production data

   weight 1 · round to HubSpot
   HubSpot✓fullclaimed7/10

   HubSpot explicitly documents the ability to create up to 10 free test/sandbox accounts to test apps and integrations without affecting real production data, directly matching the story. Missing for 10: no independent/hands-on confirmation of sandbox fidelity or limitations, and no detail on how closely sandbox mirrors production data/schema.

   • [claimed-docs] “You can create up to 10 test accounts to test apps and integrations without affecting any real HubSpot data.”

   proof ↗⚑ flag

   Twenty—none0/10

   No evidence of a sandbox/staging environment, test workspace, or non-production mode for safely testing AI agents or workflows separate from live data. Self-hosting via Docker Compose could theoretically allow spinning up a separate instance, but this is never documented as a sandbox testing feature. missing for 10: dedicated sandbox/staging environment, data seeding/test-mode tooling, documentation of testing AI agents against non-production data.

   ⚑ flag

4. ai-native userRely on versioned APIs with a documented deprecation policy

   weight 2 · round to HubSpot
   HubSpot✓fullclaimed8/10

   HubSpot documents a date-based API versioning scheme (e.g. /crm/2026-09/) where prior versions continue working until an announced end-of-life date, plus a maintained 'Sunsetted and deprecated APIs' page with per-API dates/migration targets and a commitment to announce changes via the Developer Changelog with ample notice. Each endpoint also publishes a machine-readable OpenAPI spec inline, reinforcing an AI-native/versioned-API workflow. Missing for 10: independent/third-party corroboration of the deprecation policy being honored in practice, and the top-level llms.txt/openapi.json probes returned 404 at some paths, slightly undercutting full agent-discoverability claims.

   • [claimed-docs] “Docs, "2026-09 API reference": "The 2026-09 APIs use a date-based versioning scheme (e.g., /crm/2026-09/)... This versioning model gives Hub…”
   • [claimed-docs] “Docs, "Sunsetted and deprecated APIs": a maintained table of sunsetted APIs with dates and migration targets, plus "Deprecated APIs... HubSp…”
   • [claimed-docs] “Every endpoint reference page publishes its machine-readable spec inline ("## OpenAPI ... specs/2026-09/account-account-info-v2026-09.json G…”

   proof ↗⚑ flag

   Twenty—none0/10

   Evidence documents REST/GraphQL APIs with auto-generated docs and metadata endpoints, but there is no mention of API versioning scheme or a documented deprecation policy anywhere in the pack; probe results even show openapi spec endpoints returning 404.

   • [claimed-docs] “A developer-first API that adapts to your custom data model. Both GraphQL and REST endpoints, with auto-generated documentation per works…”
   • [claimed-docs] “when you add a custom object (say Invoice), it immediately gets REST and GraphQL endpoints identical to built-in objects like Company or Per…”
   • [claimed-docs] “Metadata API — /rest/metadata/ and /metadata/ Schema management: create/modify/delete objects, fields, and relations. This is how you progra…”
   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.twenty.com/openapi.json, https://docs.twenty.com/swagger.json, https://docs.twenty.com/…”

   proof ↗⚑ flag

1. ai-native userPerform bulk operations across many items at once

   weight 2 · round to Twenty
   HubSpot~partialclaimed6/10

   HubSpot's docs confirm batch/bulk capabilities such as retrieving custom objects in batches, bulk import of multiple object records, and CRM-wide sync/dedup tools, and its MCP server gives AI agents read/write access to CRM data. However, there's no explicit documentation of bulk create/update/delete operations being exposed specifically for AI-agent workflows (e.g., no mention of batch endpoints for contacts/deals via the MCP server) and no independent verification of large-scale bulk performance. Missing for 10: explicit agent-facing bulk write/delete API examples, evidence of batch limits/performance, and independent/hands-on confirmation of bulk operations at scale.

   • [claimed-docs] “You can retrieve custom object records individually or in batches.”
   • [claimed-docs] “Import records for multiple objects to add or update data across different objects. You can import records such as companies or deals, as we…”
   • [claimed-docs] “Import records for multiple objects to add or update data across different objects.”
   • [claimed-docs] “The HubSpot MCP server enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — conta…”
   • [claimed-docs] “enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — contacts, deals, engagements…”
   • [claimed-docs] “HubSpot automatically deduplicates contacts using email addresses, and companies using domain names.”
   • [claimed-docs] “Record IDs can also be used to manually deduplicate contacts, companies, deals, tickets, products, admin-activated objects (e.g., appointmen…”

   proof ↗⚑ flag

   Twenty✓fullclaimed8/10

   Twenty's REST and GraphQL APIs explicitly support batching up to 60 records per request for create/update/delete, plus GraphQL batch upsert, and CSV/XLSX/API import with unlimited records for large-scale migrations, alongside export up to 20,000 records — directly enabling bulk operations across many items. Missing for 10: independent/hands-on verification of batch API limits in practice and no documented bulk-delete-via-filter or bulk-edit-in-UI feature beyond import/export and API batching.

   • [claimed-docs] “Both REST and GraphQL support batching up to 60 records per request — create, update, or delete. GraphQL also supports batch upsert”
   • [claimed-docs] “Import data from CSV files or via API. Field mapping, duplicate detection, and error handling built in. Export your data anytime — no loc…”
   • [claimed-docs] “API Import — Large-scale migrations, automation — Unlimited”
   • [claimed-docs] “Export your workspace data for backups, reporting, or migration. ... Export limits: Up to 20,000 records per export.”
   • [claimed-docs] “Map your columns to Twenty fields ... Review errors (highlighted in yellow) and fix them, directly editing in the UI”
   • [claimed-docs] “You can import data for any object using CSV, XLSX, or XLS files. Each file should contain only one type of object (e.g., only People record…”

   proof ↗⚑ flag

2. ai-native userDefine rules that trigger actions automatically on events

   weight 3 · round drawn
   HubSpot✓fullclaimed8/10

   HubSpot Workflows (hubspot-docs-12/45) let users define automation rules triggered by events (contact property changes, form submissions, deal stage changes, etc.), and this is complemented by webhooks/event subscription APIs (hubspot-docs-4) for programmatic event-driven automation, plus AI agents (Prospecting agent, deal progression) that act automatically on signals (hubspot-docs-14, hubspot-docs-22, hubspot-docs-23). missing for 10: no detailed documentation of conditional/branching logic specifics or independent hands-on verification of workflow trigger reliability.

   • [claimed-docs] “Create workflows in HubSpot to automate your processes.”
   • [claimed-docs] “Create workflows”
   • [claimed-docs] “the webhooks journal API enables third-party integrators to subscribe to real-time events, retrieve historical event data, and manage snapsh…”
   • [claimed-docs] “Prospecting agent watches accounts for signals that they're ready to buy and reaches out automatically, while smart deal progression keeps d…”
   • [claimed-docs] “Prospecting agent watches accounts for signals that they're ready to buy and reaches out automatically, while smart deal progression keeps d…”
   • [claimed-docs] “Approve automatic CRM updates and follow-up drafts after every meeting”

   proof ↗⚑ flag

   Twenty✓fullclaimed8/10

   Twenty's Workflows feature explicitly supports rule-based automation: triggers on record changes, schedules, manual actions, or incoming webhooks, chaining actions like record operations, HTTP requests, code, branches, AI Agent steps, Slack alerts, and enrichment — a clear rules-trigger-actions automation engine documented with concrete examples (deal stage → Slack alert, new contact → enrichment). Missing for 10: independent hands-on verification of workflow reliability/complexity limits beyond docs and one general community note about alpha-stage issues (not specific to workflows).

   • [claimed-docs] “Trigger workflows on record changes, schedules, manual actions, or incoming webhooks.”
   • [claimed-docs] “Workflows can chain any combination of: Record operations... Send email... HTTP request... Code... Branches... Iterator... AI Agent... Delay…”
   • [claimed-docs] “Send Slack alerts when a deal reaches a certain stage * Auto-enrich new contacts with data from external APIs”
   • [claimed-docs] “Workflows let you automate repetitive tasks and connect Twenty to external tools — without writing code (though you can if you want to).”
   • [claimed-docs] “Code — Run custom JavaScript for complex logic”
   • [claimed-docs] “Workflows let you automate repetitive tasks and connect Twenty to external tools — without writing code”

   proof ↗⚑ flag

3. ai-native userSchedule recurring jobs or workflows

   weight 2 · round to Twenty
   HubSpot~partialclaimed4/10

   HubSpot's Workflows feature lets users automate processes (hubspot-docs-12/45), which is the closest capability to scheduling recurring jobs, but the evidence pack gives no detail on scheduling logic, cron-like recurrence, or how an AI-native user (via MCP/API) would configure or trigger recurring workflows programmatically. Missing for 10: documentation on recurring/scheduled triggers, API/MCP support for creating or managing workflow schedules, and evidence of AI-agent-initiated recurring automation.

   • [claimed-docs] “Create workflows in HubSpot to automate your processes.”
   • [claimed-docs] “Create workflows”

   proof ↗⚑ flag

   Twenty✓fullclaimed7/10

   Twenty's workflow engine explicitly supports schedule-based triggers alongside record changes, manual actions, and webhooks, and workflows can chain multi-step actions (HTTP, code, branches, AI agent, delay) enabling recurring automation chains. Missing for 10: independent/hands-on confirmation of recurrence configuration (e.g., cron intervals) and no community validation of schedule-trigger reliability.

   • [claimed-docs] “Trigger workflows on record changes, schedules, manual actions, or incoming webhooks.”
   • [claimed-docs] “Workflows can chain any combination of: Record operations... Send email... HTTP request... Code... Branches... Iterator... AI Agent... Delay…”
   • [claimed-docs] “Workflows let you automate repetitive tasks and connect Twenty to external tools — without writing code (though you can if you want to).”

   proof ↗⚑ flag

4. ai-native userVersion, review, and roll back my automations

   weight 1 · round drawn
   HubSpot—none0/10

   The evidence confirms HubSpot workflows exist for automation (hubspot-docs-12/45), and general API versioning/deprecation policies are documented, but nothing in the pack describes workflow-level version history, change review, or rollback capability for automations themselves. Missing for 10: workflow revision history UI, diff/review of automation changes, and rollback-to-previous-version functionality.

   • [claimed-docs] “Create workflows in HubSpot to automate your processes.”
   • [claimed-docs] “Create workflows”
   • [claimed-docs] “Docs, "2026-09 API reference": "The 2026-09 APIs use a date-based versioning scheme (e.g., /crm/2026-09/)... This versioning model gives Hub…”

   proof ↗⚑ flag

   Twenty—none0/10

   The evidence describes Twenty's workflow builder (triggers, chaining actions, code steps) but contains no mention of version history, change review, or rollback capability for workflows/automations. Missing for 10: any documentation of workflow version tracking, diff/review UI, or rollback/undo mechanism for automations.

   • [claimed-docs] “Workflows can chain any combination of: Record operations... Send email... HTTP request... Code... Branches... Iterator... AI Agent... Delay…”
   • [claimed-docs] “Workflows let you automate repetitive tasks and connect Twenty to external tools — without writing code (though you can if you want to).”
   • [claimed-docs] “Code — Run custom JavaScript for complex logic”

   proof ↗⚑ flag

Workflows

1. ops userRoute new leads to the right owner automatically with assignment rules or round-robin

   weight 2 · round to HubSpot
   HubSpot~partialclaimed3/10

   HubSpot's documentation confirms a general-purpose workflow automation engine ('Create workflows in HubSpot to automate your processes') that is the underlying mechanism used for lead routing, but the evidence pack contains no explicit mention of assignment rules, lead routing, or round-robin distribution features. Missing for 10: explicit documentation of assignment rules, round-robin lead distribution, or lead-routing-specific workflow actions, and any independent/hands-on confirmation these work as described.

   • [claimed-docs] “Create workflows in HubSpot to automate your processes.”
   • [claimed-docs] “Create workflows”

   proof ↗⚑ flag

   Twenty—none0/10

   Twenty's workflow engine (record-change triggers, branches, record operations, code) could theoretically be scripted to assign owners, but there is no documented assignment-rule or round-robin lead-routing feature anywhere in the evidence pack.

   • [claimed-docs] “Trigger workflows on record changes, schedules, manual actions, or incoming webhooks.”
   • [claimed-docs] “Workflows can chain any combination of: Record operations... Send email... HTTP request... Code... Branches... Iterator... AI Agent... Delay…”
   • [claimed-docs] “Send Slack alerts when a deal reaches a certain stage * Auto-enrich new contacts with data from external APIs”

   proof ↗⚑ flag

2. ops userBuild trigger-based workflows that update records, create tasks, and send notifications automatically

   weight 3 · round drawn
   HubSpot✓fullclaimed8/10

   HubSpot's Workflows product explicitly supports trigger-based automation to update records, create tasks, and send notifications, and this is corroborated by the broader CRM automation ecosystem (deduplication, email sync, sequences) that workflows commonly integrate with. Missing for 10: no independent/hands-on account of building a multi-action trigger workflow (task+notification+record update chain), and docs excerpts are terse marketing-style snippets rather than detailed workflow-builder documentation.

   • [claimed-docs] “Create workflows in HubSpot to automate your processes.”
   • [claimed-docs] “Create workflows”
   • [claimed-docs] “HubSpot automatically deduplicates contacts using email addresses, and companies using domain names.”
   • [claimed-docs] “With the sequences tool, you can send a series of targeted, timed”

   proof ↗⚑ flag

   Twenty✓fullcommunity8/10

   Twenty's docs directly describe trigger-based workflows (record changes, schedules, manual, webhooks) that can chain record operations, send email, HTTP requests, code, branches, and specifically cite examples like Slack alerts on deal stage changes and auto-enriching contacts, matching the ops story's update/create/notify pattern. Missing for 10: independent hands-on validation of workflow reliability in production (community evidence flags general 'alpha' rough edges) and no explicit dedicated 'create task' workflow action example beyond generic record operations.

   • [claimed-docs] “Trigger workflows on record changes, schedules, manual actions, or incoming webhooks.”
   • [claimed-docs] “Workflows can chain any combination of: Record operations... Send email... HTTP request... Code... Branches... Iterator... AI Agent... Delay…”
   • [claimed-docs] “Send Slack alerts when a deal reaches a certain stage * Auto-enrich new contacts with data from external APIs”
   • [claimed-docs] “Workflows let you automate repetitive tasks and connect Twenty to external tools — without writing code (though you can if you want to).”
   • [claimed-docs] “Code — Run custom JavaScript for complex logic”
   • [community] “Twenty was almost what we needed. It was extremely easy to set up, the customization is very promising, and the REST API worked ok. The main…”

   proof ↗⚑ flag

Ai crm ops

1. ai-native userHave an agent enrich a new lead end-to-end via the API or MCP server — creating the record, filling company data, and assigning an owner

   weight 3 · round to Twenty
   HubSpot~partialprobed5/10

   HubSpot's REST API supports creating contact records (POST /crm/v3/objects/contacts) and companies/custom objects can be created and updated via the same object APIs, and the official MCP server explicitly grants 'read and write access to your HubSpot CRM data — contacts, deals, engagements, and more' to any MCP-compatible agent, so the create+enrich portions of the story are documented. However, there is no evidence of an owner-assignment field/endpoint being called out, nor any end-to-end example showing an agent orchestrating create→enrich→assign-owner in one flow. Missing for 10: explicit owner-assignment API/MCP action, a documented end-to-end lead-enrichment workflow example, and independent/hands-on confirmation of the MCP server performing writes successfully.

   • [claimed-docs] “To create one contact, make a `POST` request to `/crm/v3/objects/contacts`.”
   • [claimed-docs] “The HubSpot MCP server enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — conta…”
   • [claimed-docs] “The HubSpot MCP server enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data”
   • [claimed-docs] “enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — contacts, deals, engagements…”
   • [claimed-docs] “The contacts endpoints allow you to create and manage contact records in your HubSpot account, as well as sync contact data between HubSpot …”
   • [claimed-docs] “HubSpot custom objects allow organizations to represent and organize data based on unique business requirements. Create and update custom ob…”
   • [claimed-docs] “use the objects API to create and manage the custom object’s records”
   • [probe] “official MCP server documented at https://developers.hubspot.com/ai-tools/mcp”

   proof ↗⚑ flag

   Twenty~partialclaimed6/10

   Twenty documents an MCP server (twenty-docs-32) and AI agents that can 'enrich records with data from external sources' and execute multi-step tasks (twenty-docs-56, twenty-docs-5), plus a developer API capable of creating custom-object records (twenty-docs-8, twenty-docs-22) and workflows that 'auto-enrich new contacts with data from external APIs' (twenty-docs-14). However, there is no explicit documentation of an agent assigning a record owner as part of this flow, nor a concrete end-to-end example (lead creation + enrichment + owner assignment) via API or MCP, and no independent/hands-on confirmation that the MCP server performs this workflow. Missing for 10: explicit owner-assignment step in agent/API flow, a documented end-to-end example combining all three actions, independent verification of MCP server behavior.

   • [claimed-docs] “MCP server | Yes | Yes | Yes”
   • [claimed-docs] “AI agents go beyond chat — they can execute multi-step tasks autonomously: Enrich records with data from external sources”
   • [claimed-docs] “AI agents that work autonomously within your CRM — answering questions, enriching records, and executing multi-step tasks within your permis…”
   • [claimed-docs] “Send Slack alerts when a deal reaches a certain stage * Auto-enrich new contacts with data from external APIs”
   • [claimed-docs] “A developer-first API that adapts to your custom data model. Both GraphQL and REST endpoints, with auto-generated documentation per works…”
   • [claimed-docs] “when you add a custom object (say Invoice), it immediately gets REST and GraphQL endpoints identical to built-in objects like Company or Per…”
   • [claimed-docs] “Agents can only access objects and fields that the user (or role) has permission to view. Sensitive data stays protected even when AI is inv…”

   proof ↗⚑ flag

Data quality

1. ops userHave contact and company records auto-enriched with firmographic and contact data

   weight 2 · round to Twenty
   HubSpot—none0/10

   The evidence pack covers CRM data unification, deduplication, import, and AI agents, but contains no mention of automatic firmographic/contact enrichment (e.g., Breeze Intelligence or third-party data append) for contact and company records.

   ⚑ flag

   Twenty~partialclaimed5/10

   Docs mention enrichment only as an example built via custom Workflows (HTTP request to an external API) and via AI Agents that can 'enrich records with data from external sources,' but there's no evidence of a built-in, native firmographic/contact enrichment provider or one-click integration—it requires the ops user (or a dev) to configure a workflow or agent with a third-party data source. Missing for 10: a native/pre-built enrichment integration (e.g., built-in data provider), independent/hands-on confirmation that enrichment works as advertised, and detail on what firmographic data fields are actually populated.

   • [claimed-docs] “Send Slack alerts when a deal reaches a certain stage * Auto-enrich new contacts with data from external APIs”
   • [claimed-docs] “AI agents go beyond chat — they can execute multi-step tasks autonomously: Enrich records with data from external sources”
   • [claimed-docs] “AI agents that work autonomously within your CRM — answering questions, enriching records, and executing multi-step tasks within your permis…”
   • [claimed-docs] “Agents can only access objects and fields that the user (or role) has permission to view. Sensitive data stays protected even when AI is inv…”

   proof ↗⚑ flag

2. ops userDetect duplicate contacts or companies and merge them without losing data

   weight 2 · round to HubSpot
   HubSpot✓fullclaimed7/10

   HubSpot docs confirm automatic deduplication (by email for contacts, domain for companies) plus manual merge via record IDs across contacts, companies, deals, tickets and other objects, directly addressing duplicate detection and merging. Missing for 10: explicit documentation confirming no data loss during merge (e.g., how conflicting field values are preserved) and independent/hands-on confirmation of merge behavior beyond vendor docs.

   • [claimed-docs] “HubSpot automatically deduplicates contacts using email addresses, and companies using domain names.”
   • [claimed-docs] “Record IDs can also be used to manually deduplicate contacts, companies, deals, tickets, products, admin-activated objects (e.g., appointmen…”
   • [claimed-docs] “HubSpot automatically deduplicates contacts using email addresses, and companies using domain names. Record IDs can also be used to manually…”
   • [claimed-docs] “Deduplicate records in HubSpot”

   proof ↗⚑ flag

   Twenty~partialclaimed4/10

   Docs confirm duplicate detection during CSV/API import (twenty-docs-9), but there is no evidence of a dedicated merge tool or workflow for combining existing duplicate Contact/Company records without losing data — only prevention-at-import is documented. missing for 10: explicit merge UI/feature for existing duplicates, data-preservation guarantees during merge, independent/hands-on confirmation of merge behavior.

   • [claimed-docs] “Import data from CSV files or via API. Field mapping, duplicate detection, and error handling built in. Export your data anytime — no loc…”
   • [claimed-docs] “Map your columns to Twenty fields ... Review errors (highlighted in yellow) and fix them, directly editing in the UI”

   proof ↗⚑ flag

Record management

1. ops userManage contacts and companies as linked records with a full activity timeline on each

   weight 3 · round drawn
   HubSpot~partialclaimed6/10

   HubSpot's docs confirm contacts and companies exist as CRM records with dedup by domain (companies) and email (contacts), the ability to import and associate records across objects (contacts, companies, deals), and email/replies being logged to the CRM — indicating activity capture on records. However, no direct evidence describes a company-record API or a rendered 'activity timeline' UI feature explicitly, only inferred from email-logging and engagement mentions in MCP docs. Missing for 10: explicit documentation of the company object API/endpoints, and explicit description of a unified activity timeline UI on contact/company records.

   • [claimed-docs] “Import records for multiple objects to add or update data across different objects. You can import records such as companies or deals, as we…”
   • [claimed-docs] “HubSpot automatically deduplicates contacts using email addresses, and companies using domain names.”
   • [claimed-docs] “Connect an individual work email to HubSpot to send one-to-one emails from the CRM, log email replies to the CRM, send sequences emails, and…”
   • [claimed-docs] “Connect an individual work email to HubSpot to send one-to-one emails from the CRM, log email replies to the CRM, send sequences emails”
   • [claimed-docs] “To create one contact, make a `POST` request to `/crm/v3/objects/contacts`.”
   • [claimed-docs] “enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — contacts, deals, engagements…”

   proof ↗⚑ flag

   Twenty~partialcommunity6/10

   Twenty's docs confirm core CRM objects (Company/Person) with custom relations, email/calendar auto-linking to records, and workflow/API support, which supports linked-record management, but there is no explicit documentation of a unified 'activity timeline' feature showing notes, emails, calls, and tasks chronologically per record. Community evidence is limited to general alpha-stage caveats, not specific timeline testing. Missing for 10: explicit documentation or screenshots of an activity timeline UI per record, independent/hands-on verification of timeline behavior, and confirmation of what activity types (calls, notes, tasks) populate it.

   • [claimed-docs] “You can create custom relations between any objects, including many-to-many relationships.”
   • [claimed-docs] “Connect your **Google Workspace** or **Microsoft 365** mailbox. Once connected: Emails are automatically linked to the matching Company and …”
   • [claimed-docs] “Emails are automatically linked to the matching Company and People records”
   • [claimed-docs] “Emails and calendar events automatically appear on the relevant CRM records.”
   • [community] “Twenty was almost what we needed. It was extremely easy to set up, the customization is very promising, and the REST API worked ok. The main…”

   proof ↗⚑ flag

Data model

1. ops userAdd custom fields and build filtered, sorted, and shared views without an admin project

   weight 2 · round to Twenty
   HubSpot—none0/10

   The evidence covers custom objects via the developer API, reporting dashboards with filters, and workflow automation, but there is no evidence of an ops user creating custom fields/properties or building filtered, sorted, shared record list views through the UI without engineering/admin involvement — the closest material (custom objects API, reporting dashboard filters) targets developers or reporting, not object-view customization.

   • [claimed-docs] “HubSpot custom objects allow organizations to represent and organize data based on unique business requirements. Create and update custom ob…”
   • [claimed-docs] “Create and update custom object records with the custom objects API.”
   • [claimed-docs] “Build custom reporting dashboards for every team”
   • [claimed-docs] “Build marketing, sales, revenue, and service analytics reports with filters and chart types to see exactly what's happening across your busi…”

   proof ↗⚑ flag

   Twenty✓fullcommunity8/10

   Docs clearly show custom objects/fields (20+ types, unlimited custom objects), table/kanban/calendar views with AND/OR filtering, multi-field sort, grouping, and saved custom views — all configurable in-app without dev involvement, plus community corroboration that customization was 'very promising' and easy to set up. Missing for 10: explicit documentation of 'sharing' a saved view with specific teammates/permissions (only role-based permissions and workspace-level dashboard sharing are documented, not view-sharing specifics) and independent hands-on confirmation of the filter/sort/save-view workflow itself.

   • [claimed-docs] “Create custom objects, add custom fields with 20+ field types, and build relationships between any objects.”
   • [claimed-docs] “Switch between table views, kanban boards, and calendar views. Filter with AND/OR logic, sort by multiple fields, group records, and save cu…”
   • [claimed-docs] “You can create custom relations between any objects, including many-to-many relationships.”
   • [claimed-docs] “You can also create custom objects for anything your business needs — projects, support tickets, products, contracts, or anything else.”
   • [claimed-docs] “Custom objects get the same first-class treatment as built-in ones — including API endpoints, views, permissions, and workflow triggers.”
   • [community] “Twenty was almost what we needed. It was extremely easy to set up, the customization is very promising, and the REST API worked ok. The main…”

   proof ↗⚑ flag

2. developerModel custom objects with fields and relationships beyond the built-in contacts, companies, and deals

   weight 3 · round to Twenty
   HubSpot✓fullclaimed7/10

   HubSpot's docs explicitly cover custom objects: creating/updating records via the custom objects API and retrieving them individually or in batch, letting developers extend beyond contacts/companies/deals. Evidence also shows import/dedup/association support extending to custom objects. Missing for 10: explicit documentation of defining custom fields/properties and relationship/association types on custom objects, and independent hands-on developer corroboration beyond HubSpot's own docs.

   • [claimed-docs] “HubSpot custom objects allow organizations to represent and organize data based on unique business requirements. Create and update custom ob…”
   • [claimed-docs] “You can retrieve custom object records individually or in batches.”
   • [claimed-docs] “use the objects API to create and manage the custom object’s records”
   • [claimed-docs] “Create and update custom object records with the custom objects API.”
   • [claimed-docs] “Record IDs can also be used to manually deduplicate contacts, companies, deals, tickets, products, admin-activated objects (e.g., appointmen…”

   proof ↗⚑ flag

   Twenty✓fullcommunity9/10

   Twenty's docs extensively cover creating custom objects, custom fields (20+ types), and building custom relations (including many-to-many) between any objects, with custom objects receiving first-class treatment (API endpoints, views, permissions, workflows) identical to built-ins. This is corroborated by hands-on community feedback praising the customization as 'very promising.' Missing for 10: deeper independent/hands-on verification of complex relationship modeling beyond a single HN mention, and no third-party technical review of the metadata/schema API in practice.

   • [claimed-docs] “Create custom objects, add custom fields with 20+ field types, and build relationships between any objects.”
   • [claimed-docs] “You can create custom relations between any objects, including many-to-many relationships.”
   • [claimed-docs] “Custom objects get the same first-class treatment as built-in ones — including API endpoints, views, permissions, and workflow triggers.”
   • [claimed-docs] “You can also create custom objects for anything your business needs — projects, support tickets, products, contracts, or anything else.”
   • [claimed-docs] “when you add a custom object (say Invoice), it immediately gets REST and GraphQL endpoints identical to built-in objects like Company or Per…”
   • [claimed-docs] “Metadata API — /rest/metadata/ and /metadata/ Schema management: create/modify/delete objects, fields, and relations. This is how you progra…”
   • [community] “Twenty was almost what we needed. It was extremely easy to set up, the customization is very promising, and the REST API worked ok. The main…”

   proof ↗⚑ flag

Email

1. ops userSync my calendar so meetings appear on contact and deal timelines

   weight 2 · round to Twenty
   HubSpot—none0/10

   The evidence pack covers email inbox connection, contact/deal APIs, imports, and dedup, but nowhere documents a calendar-sync feature that logs meetings onto contact/deal timelines. Without explicit evidence of calendar integration, this applicable CRM capability cannot be credited.

   ⚑ flag

   Twenty✓fullclaimed8/10

   Docs describe connecting Google Workspace/Microsoft 365 accounts so calendar events automatically appear on relevant CRM records (contacts/deals), with filtering controls; calendar view is also supported in record timelines. missing for 10: no independent/hands-on confirmation that calendar sync specifically (vs. email sync) populates deal timelines, and no explicit mention of deal-object calendar linking beyond Company/People.

   • [claimed-docs] “Connect your Google Workspace or Microsoft 365 account. Emails and calendar events automatically appear on the relevant CRM records.”
   • [claimed-docs] “Emails and calendar events automatically appear on the relevant CRM records.”
   • [claimed-docs] “You control what gets imported — filter by date range or sender to avoid pulling in irrelevant emails.”
   • [claimed-docs] “Connect your **Google Workspace** or **Microsoft 365** mailbox. Once connected: Emails are automatically linked to the matching Company and …”
   • [claimed-docs] “Emails are automatically linked to the matching Company and People records”

   proof ↗⚑ flag

2. ops userConnect Gmail or Outlook so emails sync two-way and log automatically to the right records

   weight 3 · round to Twenty
   HubSpot~partialclaimed6/10

   Docs confirm connecting an individual work email (Gmail/Outlook) to log replies and enable one-to-one email sync to CRM records, which addresses the core of the story. However, evidence doesn't explicitly confirm two-way sync (both sending and receiving auto-logged to right records) or details on matching/logging accuracy, contact-level automation nuances, or independent/hands-on verification. missing for 10: explicit two-way sync confirmation, details on automatic record-matching logic, independent/hands-on verification of reliability.

   • [claimed-docs] “Connect an individual work email to HubSpot to send one-to-one emails from the CRM, log email replies to the CRM, send sequences emails, and…”
   • [claimed-docs] “Connect an individual work email to HubSpot to send one-to-one emails from the CRM, log email replies to the CRM, send sequences emails”
   • [claimed-docs] “Connect an individual work email to HubSpot to send one-to-one emails from the CRM, log email replies to the CRM”
   • [claimed-docs] “Connect an individual work email”

   proof ↗⚑ flag

   Twenty✓fullclaimed8/10

   Twenty's docs explicitly describe connecting Google Workspace or Microsoft 365 accounts so emails and calendar events automatically appear on and link to matching Company/People CRM records, with filters to control what gets imported. This is documented first-party functionality directly matching the story, though there is no independent/hands-on confirmation of two-way sync reliability specifically for email. Missing for 10: independent or hands-on verification of the email sync feature itself (the community evidence found only mentions general alpha rough edges, not this feature specifically).

   • [claimed-docs] “Connect your Google Workspace or Microsoft 365 account. Emails and calendar events automatically appear on the relevant CRM records.”
   • [claimed-docs] “You control what gets imported — filter by date range or sender to avoid pulling in irrelevant emails.”
   • [claimed-docs] “Connect your **Google Workspace** or **Microsoft 365** mailbox. Once connected: Emails are automatically linked to the matching Company and …”
   • [claimed-docs] “Emails are automatically linked to the matching Company and People records”
   • [claimed-docs] “Emails and calendar events automatically appear on the relevant CRM records.”

   proof ↗⚑ flag

Outreach

1. ops userSend templated emails or multi-step sequences to prospects from inside the CRM

   weight 2 · round to HubSpot
   HubSpot✓fullclaimed8/10

   HubSpot docs clearly document connecting an inbox to send one-to-one emails from the CRM and a dedicated sequences tool for sending multi-step, timed email sequences to prospects, both directly accessible from within the CRM. This directly matches the ops persona's need for templated/sequence outreach. Missing for 10: no independent/hands-on corroboration of the sequences UX and no explicit mention of email templates feature alongside sequences.

   • [claimed-docs] “Connect an individual work email to HubSpot to send one-to-one emails from the CRM, log email replies to the CRM, send sequences emails, and…”
   • [claimed-docs] “With the sequences tool, you can send a series of targeted, timed”
   • [claimed-docs] “With the sequences tool, you can send a series of targeted, timed [email temp”
   • [claimed-docs] “Create and edit sequences”
   • [claimed-docs] “Connect an individual work email to HubSpot to send one-to-one emails from the CRM, log email replies to the CRM, send sequences emails”

   proof ↗⚑ flag

   Twenty~partialclaimed4/10

   Twenty's workflow engine can chain a 'Send email' step with Delay and Branch steps, which could technically be assembled into a multi-step email sequence, and email sync auto-links messages to CRM records — but there is no documented email-template feature, no dedicated 'sequence/cadence' tool for prospects, and no hands-on evidence this is actually used this way. Missing for 10: dedicated templated-email/sequence builder for outbound prospecting, evidence of template management, and independent confirmation of ops teams using workflows this way.

   • [claimed-docs] “Workflows can chain any combination of: Record operations... Send email... HTTP request... Code... Branches... Iterator... AI Agent... Delay…”
   • [claimed-docs] “Code — Run custom JavaScript for complex logic”
   • [claimed-docs] “Connect your **Google Workspace** or **Microsoft 365** mailbox. Once connected: Emails are automatically linked to the matching Company and …”
   • [claimed-docs] “Emails are automatically linked to the matching Company and People records”

   proof ↗⚑ flag

Api depth

1. developerSync records bidirectionally through the API at scale, with documented rate limits, upserts, and change detection

   weight 3 · round drawn
   HubSpot~partialclaimed5/10

   Evidence shows CRUD and batch operations for contacts/custom objects, plus a webhooks journal API for real-time and historical event subscription (a change-detection mechanism), supporting bidirectional sync at scale. However, no evidence cites documented rate limits, and there's no explicit mention of 'upsert' semantics (e.g., idProperty-based upsert) in the API docs provided. Missing for 10: documented rate-limit numbers/headers, explicit upsert endpoint documentation, and independent/hands-on confirmation of scale performance.

   • [claimed-docs] “You can retrieve custom object records individually or in batches.”
   • [claimed-docs] “the webhooks journal API enables third-party integrators to subscribe to real-time events, retrieve historical event data, and manage snapsh…”
   • [claimed-docs] “The contacts endpoints allow you to create and manage contact records in your HubSpot account, as well as sync contact data between HubSpot …”
   • [claimed-docs] “HubSpot custom objects allow organizations to represent and organize data based on unique business requirements. Create and update custom ob…”
   • [claimed-docs] “use the objects API to create and manage the custom object’s records”

   proof ↗⚑ flag

   Twenty~partialcommunity5/10

   Docs show REST/GraphQL batching (up to 60 records) with GraphQL batch upsert, a Metadata API, and workflow triggers on record changes/webhooks that could support change detection, plus import/export capabilities for large-scale sync. However, no documented API rate limits are provided anywhere in the evidence, and community feedback flags the REST API as 'alpha' with unspecified issues, undercutting confidence in scale reliability. missing for 10: explicit documented rate limits, dedicated change-detection/webhook-for-sync documentation beyond generic workflow triggers, independent verification of sync at scale.

   • [claimed-docs] “Both REST and GraphQL support batching up to 60 records per request — create, update, or delete. GraphQL also supports batch upsert”
   • [claimed-docs] “Metadata API — /rest/metadata/ and /metadata/ Schema management: create/modify/delete objects, fields, and relations. This is how you progra…”
   • [claimed-docs] “API Import — Large-scale migrations, automation — Unlimited”
   • [claimed-docs] “Export your workspace data for backups, reporting, or migration. ... Export limits: Up to 20,000 records per export.”
   • [claimed-docs] “Trigger workflows on record changes, schedules, manual actions, or incoming webhooks.”
   • [community] “Twenty was almost what we needed. It was extremely easy to set up, the customization is very promising, and the REST API worked ok. The main…”

   proof ↗⚑ flag

Ecosystem

1. ops userInstall prebuilt integrations (Slack, support desk, marketing, data warehouse) from an app marketplace

   weight 2 · round to Twenty
   HubSpot—none0/10

   The evidence pack covers HubSpot's CRM APIs, CLI, MCP servers, import tools, and AI features, but contains no mention of an App Marketplace or prebuilt one-click integrations for Slack, support desks, marketing tools, or data warehouses. While this axis clearly applies to a SaaS platform like HubSpot, there is no evidence in the pack that such a marketplace or prebuilt connector catalog exists or is documented.

   ⚑ flag

   Twenty~partialclaimed4/10

   Twenty supports connecting to external tools via Zapier (5,000+ apps) and lets developers build/publish apps for the community, plus workflow actions like Slack alerts, but there's no evidence of a curated app marketplace with one-click prebuilt integrations specifically for Slack, support desk, marketing, or data warehouse tools — these have to be built as workflows or via Zapier rather than installed as packaged apps. Missing for 10: a dedicated marketplace UI, named prebuilt connectors for support desk/marketing/warehouse categories, and evidence of one-click install rather than custom workflow/code setup.

   • [claimed-docs] “Zapier | Connect to 5,000+ apps without code”
   • [claimed-docs] “Send Slack alerts when a deal reaches a certain stage * Auto-enrich new contacts with data from external APIs”
   • [claimed-docs] “Developers can build full-featured apps on top of Twenty — adding custom UI, server-side logic, and deep integrations. Apps can be published…”
   • [github] “npx twenty app:publish --private”
   • [github] “npx create-twenty-app my-app ... npx twenty app:publish --private”

   proof ↗⚑ flag

Imports

1. ops userMigrate from another CRM with my records, owners, and history intact

   weight 2 · round to HubSpot
   HubSpot~partialclaimed6/10

   HubSpot documents import tools for multiple objects (companies, deals, activities, associations), self-service transfer/audit-and-sync for new customers, and automatic/manual deduplication using email/domain or record IDs, which together support migrating records and relationships from another CRM. However, evidence does not confirm preservation of 'owners' (rep/user assignment mapping) during migration, nor explicit historical activity/timeline preservation beyond basic activities import, and there's no independent/hands-on account confirming migration fidelity. missing for 10: evidence of owner/user mapping during import, confirmation of full historical timeline/audit trail preservation, and independent verification of migration outcomes.

   • [claimed-docs] “Import records for multiple objects to add or update data across different objects. You can import records such as companies or deals, as we…”
   • [claimed-docs] “If you're new to HubSpot and want to transfer your existing CRM data to HubSpot's CRM, you can use self-service transfer to audit and sync y…”
   • [claimed-docs] “HubSpot automatically deduplicates contacts using email addresses, and companies using domain names.”
   • [claimed-docs] “Record IDs can also be used to manually deduplicate contacts, companies, deals, tickets, products, admin-activated objects (e.g., appointmen…”
   • [claimed-docs] “you can use self-service transfer to audit and sync your data”
   • [claimed-docs] “Import records for multiple objects to add or update data across different objects.”

   proof ↗⚑ flag

   Twenty~partialclaimed5/10

   Twenty documents CSV/XLSX/API import with field mapping, duplicate detection, and error handling, plus relation-matching via unique fields (id, email, domain) to preserve links between records like Company/Person — covering the core 'records and relationships' part of migration. However, there's no documented mechanism for preserving record owners (assigning imported records to specific users) or historical activity/timeline data during import, and no case study of an actual CRM-to-CRM migration succeeding end-to-end. Missing for 10: explicit owner-field mapping/preservation during import, migration of activity history/notes/timestamps, and independent evidence of a successful full CRM migration.

   • [claimed-docs] “Map your columns to Twenty fields ... Review errors (highlighted in yellow) and fix them, directly editing in the UI”
   • [claimed-docs] “You need to reference the related object using a unique field from this object: the id, the email for People and Workspace Members, the doma…”
   • [claimed-docs] “You can import data for any object using CSV, XLSX, or XLS files. Each file should contain only one type of object (e.g., only People record…”
   • [claimed-docs] “You can import data for any object using CSV, XLSX, or XLS files.”
   • [claimed-docs] “API Import — Large-scale migrations, automation — Unlimited”
   • [claimed-docs] “Import data from CSV files or via API. Field mapping, duplicate detection, and error handling built in. Export your data anytime — no loc…”

   proof ↗⚑ flag

2. founderImport my spreadsheet of contacts and deals with field mapping and validation, without losing rows silently

   weight 3 · round to Twenty
   HubSpot~partialclaimed5/10

   HubSpot documents import for multiple objects (contacts, companies, deals) with association mapping and automatic/manual deduplication to prevent duplicate rows, plus self-service transfer/audit tooling for new users. However, the evidence does not describe field-mapping UI specifics, explicit validation/error-row reporting, or confirmation that rows are never silently dropped during import. missing for 10: documented field-mapping interface details, explicit row-level validation/error reporting during import, independent/hands-on confirmation that no rows are silently lost.

   • [claimed-docs] “Import records for multiple objects to add or update data across different objects. You can import records such as companies or deals, as we…”
   • [claimed-docs] “If you're new to HubSpot and want to transfer your existing CRM data to HubSpot's CRM, you can use self-service transfer to audit and sync y…”
   • [claimed-docs] “HubSpot automatically deduplicates contacts using email addresses, and companies using domain names.”
   • [claimed-docs] “Record IDs can also be used to manually deduplicate contacts, companies, deals, tickets, products, admin-activated objects (e.g., appointmen…”
   • [claimed-docs] “you can use self-service transfer to audit and sync your data”

   proof ↗⚑ flag

   Twenty✓fullclaimed8/10

   Docs describe CSV/XLSX/XLS import with column-to-field mapping, duplicate detection, and an error-review UI that highlights bad rows in yellow for in-place fixing rather than silent dropping, plus API-based bulk import for larger migrations. Missing for 10: independent/hands-on verification that no rows are silently lost during real-world large or messy imports, and no detail on what happens to unresolved/invalid rows beyond the UI flagging step.

   • [claimed-docs] “Import data from CSV files or via API. Field mapping, duplicate detection, and error handling built in. Export your data anytime — no loc…”
   • [claimed-docs] “Map your columns to Twenty fields ... Review errors (highlighted in yellow) and fix them, directly editing in the UI”
   • [claimed-docs] “You need to reference the related object using a unique field from this object: the id, the email for People and Workspace Members, the doma…”
   • [claimed-docs] “You can import data for any object using CSV, XLSX, or XLS files. Each file should contain only one type of object (e.g., only People record…”
   • [claimed-docs] “You can import data for any object using CSV, XLSX, or XLS files.”
   • [claimed-docs] “API Import — Large-scale migrations, automation — Unlimited”

   proof ↗⚑ flag

Time to value

1. founderGo from signup to a working pipeline with my real data in under an hour, without an implementation partner

   weight 2 · round drawn
   HubSpot~partialclaimed6/10

   HubSpot documents a self-service path — free CRM signup with no expiration, self-service data transfer/import for multiple objects, automatic deduplication, and workflow creation — all achievable without a partner (hubspot-docs-7,9,25,27,29,8,10,12). However, no evidence quantifies time-to-value or confirms a real founder can reach a 'working pipeline with real data' inside one hour; community commentary only addresses marketing/product perception, not onboarding speed. Missing for 10: a documented or independent time-to-pipeline benchmark, and hands-on confirmation that import+dedup+workflow setup is fast enough for non-technical founders without support.

   • [claimed-docs] “HubSpot's free CRM unifies all your customer data on one platform, with AI that makes it easy to understand. And unlike a free trial, it's 1…”
   • [claimed-docs] “If you're new to HubSpot and want to transfer your existing CRM data to HubSpot's CRM, you can use self-service transfer to audit and sync y…”
   • [claimed-docs] “Your data deserves better than spreadsheets. HubSpot's free CRM unifies all your customer data on one platform, with AI that makes it easy t…”
   • [claimed-docs] “you can use self-service transfer to audit and sync your data”
   • [claimed-docs] “Import records for multiple objects to add or update data across different objects. You can import records such as companies or deals, as we…”
   • [claimed-docs] “HubSpot automatically deduplicates contacts using email addresses, and companies using domain names.”
   • [claimed-docs] “Create workflows in HubSpot to automate your processes.”
   • [claimed-docs] “When you’re ready for more, upgrade to get premium CRM functionality that builds on what you already have. No starting over, no data migrati…”

   proof ↗⚑ flag

   Twenty~partialcommunity6/10

   Docs show fast signup ('spin up a workspace in under a minute'), CSV/XLSX/API import with field mapping and dedup, quick custom-object/pipeline setup via kanban/table views, and one-command self-host — all consistent with rapid, self-serve onboarding without a partner. However, there's no explicit case study or benchmark confirming an actual 'under one hour to working pipeline with real data' outcome, and a hands-on HN review notes Twenty felt 'pretty alpha-version' with small issues that prevented it from fully working for their use case, tempering confidence in a frictionless first hour. Missing for 10: a concrete time-boxed onboarding case study/testimonial, and independent confirmation that data import/pipeline setup reliably completes in under an hour without hitting the rough edges the community mentioned.

   • [github] “Sign up at twenty.com and spin up a workspace in under a minute, with no infrastructure to manage and always up to date.”
   • [claimed-docs] “Import data from CSV files or via API. Field mapping, duplicate detection, and error handling built in. Export your data anytime — no loc…”
   • [claimed-docs] “Map your columns to Twenty fields ... Review errors (highlighted in yellow) and fix them, directly editing in the UI”
   • [claimed-docs] “You can import data for any object using CSV, XLSX, or XLS files. Each file should contain only one type of object (e.g., only People record…”
   • [claimed-docs] “Switch between table views, kanban boards, and calendar views. Filter with AND/OR logic, sort by multiple fields, group records, and save cu…”
   • [claimed-docs] “Install the latest stable version of Twenty with a single command: bash <(curl -sL https://raw.githubusercontent.com/twentyhq/twenty/main/pa…”
   • [community] “Twenty was almost what we needed. It was extremely easy to set up, the customization is very promising, and the REST API worked ok. The main…”

   proof ↗⚑ flag

1. ai-native userDo everything through the API that I can do in the UI

   weight 2 · round to Twenty
   HubSpot~partialprobed6/10

   HubSpot documents a broad, versioned REST API covering core CRM objects, custom objects, contacts, account info, and webhooks (hubspot-docs-2/3/4/17/18/31/32/41/57, hubspot-supp-api-versioning, hubspot-supp-openapi-spec), suggesting solid API/UI parity for core CRM functions. However, there's no evidence that newer UI-surfaced capabilities (Breeze AI reporting builder, prospecting/deal agents, sequences, workflow builder) are exposed via API, and basic agent-discoverability probes for llms.txt/docs.md/top-level openapi.json all 404 (hubspot-probe-1/2/3), undercutting a claim of full UI-API parity for AI-native consumption. missing for 10: documented API endpoints for AI agent features (prospecting agent, Breeze reporting, sequences, workflows), a working top-level machine-readable spec/llms.txt, and independent confirmation of full UI-API feature parity.

   • [claimed-docs] “the contacts endpoints allow you to create and manage contact records in your HubSpot account, as well as sync contact data between HubSpot …”
   • [claimed-docs] “You can retrieve custom object records individually or in batches.”
   • [claimed-docs] “the webhooks journal API enables third-party integrators to subscribe to real-time events, retrieve historical event data, and manage snapsh…”
   • [claimed-docs] “The contacts endpoints allow you to create and manage contact records in your HubSpot account, as well as sync contact data between HubSpot …”
   • [claimed-docs] “To create one contact, make a `POST` request to `/crm/v3/objects/contacts`.”
   • [claimed-docs] “Create and update custom object records with the custom objects API.”
   • [claimed-docs] “Retrieve detailed information about the account associated with the HubSpot portal.”
   • [claimed-docs] “Docs, "2026-09 API reference": "The 2026-09 APIs use a date-based versioning scheme (e.g., /crm/2026-09/)... This versioning model gives Hub…”
   • [claimed-docs] “Every endpoint reference page publishes its machine-readable spec inline ("## OpenAPI ... specs/2026-09/account-account-info-v2026-09.json G…”
   • [probe] “PROBE llms.txt: HTTP 404 at https://developers.hubspot.com/llms.txt”
   • [probe] “PROBE docs-md: HTTP 404 at https://developers.hubspot.com/docs.md”
   • [probe] “PROBE openapi: all candidate paths 404 (https://developers.hubspot.com/openapi.json, https://developers.hubspot.com/swagger.json, https://de…”
   • [claimed-docs] “Prospecting agent watches accounts for signals that they're ready to buy and reaches out automatically, while smart deal progression keeps d…”
   • [claimed-docs] “describe the report you need in plain language, and HubSpot will build it from your actual data.”
   • [claimed-docs] “Create and edit sequences”

   proof ↗⚑ flag

   Twenty~partialcommunity7/10

   Docs claim strong API parity — custom objects get identical REST/GraphQL endpoints, views, permissions, and workflow triggers as built-in objects, plus a Metadata API to programmatically change the data model itself, and API-based imports/exports (twenty-docs-8,12,22,23,41,59). However, no evidence explicitly confirms full API control over dashboards, view configuration (kanban/calendar), or workflow authoring — these are described only as UI-driven, and one community report notes the REST API 'worked ok' but the product was 'pretty alpha-version' with unspecified small issues. missing for 10: explicit docs/proof that dashboards, saved views, and workflow definitions can be created/edited via API rather than only through the UI, and independent confirmation beyond one mixed community report.

   • [claimed-docs] “A developer-first API that adapts to your custom data model. Both GraphQL and REST endpoints, with auto-generated documentation per works…”
   • [claimed-docs] “Custom objects get the same first-class treatment as built-in ones — including API endpoints, views, permissions, and workflow triggers.”
   • [claimed-docs] “when you add a custom object (say Invoice), it immediately gets REST and GraphQL endpoints identical to built-in objects like Company or Per…”
   • [claimed-docs] “Metadata API — /rest/metadata/ and /metadata/ Schema management: create/modify/delete objects, fields, and relations. This is how you progra…”
   • [claimed-docs] “when you add a custom object (say `Invoice`), it immediately gets REST and GraphQL endpoints identical to built-in objects like `Company` or…”
   • [claimed-docs] “API Import — Large-scale migrations, automation — Unlimited”
   • [community] “Twenty was almost what we needed. It was extremely easy to set up, the customization is very promising, and the REST API worked ok. The main…”

   proof ↗⚑ flag

2. ai-native userExport all of my data in open formats and leave

   weight 3 · round to Twenty
   HubSpot~partialclaimed3/10

   HubSpot's CRM API lets users retrieve contacts and custom objects individually or in batches, and the 'import-and-export' knowledge base implies an export path, but no evidence explicitly documents a full-account data export tool, supported open formats (CSV/JSON dump), or a process for fully leaving with all data intact. The self-service transfer feature described (hubspot-docs-9/27/29) is inbound-only (importing into HubSpot from elsewhere), not outbound. missing for 10: dedicated bulk 'export all data' feature, explicit open-format export documentation (CSV/JSON), evidence of account closure/data portability workflow, independent confirmation of successful full data export.

   • [claimed-docs] “the contacts endpoints allow you to create and manage contact records in your HubSpot account, as well as sync contact data between HubSpot …”
   • [claimed-docs] “You can retrieve custom object records individually or in batches.”
   • [claimed-docs] “Import records for multiple objects to add or update data across different objects. You can import records such as companies or deals, as we…”
   • [claimed-docs] “If you're new to HubSpot and want to transfer your existing CRM data to HubSpot's CRM, you can use self-service transfer to audit and sync y…”
   • [claimed-docs] “The contacts endpoints allow you to create and manage contact records in your HubSpot account, as well as sync contact data between HubSpot …”
   • [claimed-docs] “HubSpot custom objects allow organizations to represent and organize data based on unique business requirements. Create and update custom ob…”
   • [claimed-docs] “Import records for multiple objects to add or update data across different objects.”

   proof ↗⚑ flag

   Twenty✓fullclaimed8/10

   Twenty explicitly documents export of workspace data for backups/migration with 'no lock-in', supports CSV/XLSX export/import, self-hosting via Docker, and open REST/GraphQL APIs for full data access, giving AI-native users a clear path to extract data in open formats and leave. Missing for 10: independent hands-on verification of export completeness/fidelity and no mention of export format limits beyond CSV (e.g., no JSON/SQL dump documented) or corroboration outside vendor docs.

   • [claimed-docs] “Import data from CSV files or via API. Field mapping, duplicate detection, and error handling built in. Export your data anytime — no loc…”
   • [claimed-docs] “Export your workspace data for backups, reporting, or migration. ... Export limits: Up to 20,000 records per export.”
   • [claimed-docs] “Export your data anytime — no lock-in.”
   • [claimed-docs] “You can import data for any object using CSV, XLSX, or XLS files. Each file should contain only one type of object (e.g., only People record…”
   • [claimed-docs] “A developer-first API that adapts to your custom data model. Both GraphQL and REST endpoints, with auto-generated documentation per works…”
   • [claimed-docs] “Run Twenty on your own infrastructure with a single Docker Compose command.”
   • [claimed-docs] “Privacy-conscious organizations — you need to self-host and own your data end to end. Regulatory, contractual, or just principle.”

   proof ↗⚑ flag

3. ai-native userSelf-host the core product

   weight 3 · round to Twenty
   HubSpot—none0/10

   HubSpot is a SaaS-only CRM with no evidence of any self-hostable/open-source core; all docs describe cloud APIs, CLI, and MCP integration against hosted HubSpot accounts, not a deployable open-source package.

   ⚑ flag

   Twenty✓fullcommunity9/10

   Twenty provides clear first-party self-hosting docs, a single Docker Compose install command, and even an install script, with docs explicitly framing self-hosting as a use case for privacy-conscious organizations wanting to own their data end to end; community evidence corroborates it as easy to set up. Missing for 10: independent hands-on verification of the full self-hosted core feature set (AI, workflows) at parity with cloud beyond initial setup reports.

   • [claimed-docs] “Run Twenty on your own infrastructure with a single Docker Compose command.”
   • [claimed-docs] “Install the latest stable version of Twenty with a single command: bash <(curl -sL https://raw.githubusercontent.com/twentyhq/twenty/main/pa…”
   • [claimed-docs] “This guide provides step-by-step instructions to install and configure the Twenty application using Docker Compose.”
   • [claimed-docs] “Privacy-conscious organizations — you need to self-host and own your data end to end. Regulatory, contractual, or just principle.”
   • [claimed-docs] “**Privacy-conscious organizations** — you need to self-host and own your data end to end.”
   • [community] “Twenty was almost what we needed. It was extremely easy to set up, the customization is very promising, and the REST API worked ok. The main…”

   proof ↗⚑ flag

Ai crm ops

1. ai-native userHave an agent update deal stages and next steps from email and meeting signals via the API or MCP server

   weight 3 · round to HubSpot
   HubSpot✓fullprobed8/10

   HubSpot ships both a native AI capability ('smart deal progression keeps deals moving after every call, with next steps handled for them' and 'Approve automatic CRM updates and follow-up drafts after every meeting') and an official MCP server giving 'any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — contacts, deals, engagements, and more,' directly matching the story of an agent updating deal stages/next steps from email and meeting signals via API or MCP. Missing for 10: explicit deals-object API endpoint documentation (evidence only details contacts/custom-object endpoints), and independent hands-on confirmation that MCP-driven deal-stage updates work as advertised in practice.

   • [claimed-docs] “The HubSpot MCP server enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — conta…”
   • [claimed-docs] “Prospecting agent watches accounts for signals that they're ready to buy and reaches out automatically, while smart deal progression keeps d…”
   • [claimed-docs] “Prospecting agent watches accounts for signals that they're ready to buy and reaches out automatically, while smart deal progression keeps d…”
   • [claimed-docs] “Approve automatic CRM updates and follow-up drafts after every meeting”
   • [claimed-docs] “The HubSpot MCP server enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data”
   • [probe] “official MCP server documented at https://developers.hubspot.com/ai-tools/mcp”

   proof ↗⚑ flag

   Twenty~partialclaimed5/10

   Twenty documents the necessary building blocks — an MCP server (twenty-docs-32), REST/GraphQL API with full CRUD on custom fields like deal stage (twenty-docs-8, twenty-docs-22), automatic email/calendar linkage to CRM records (twenty-docs-4, twenty-docs-52, twenty-docs-58), and AI agents that 'execute multi-step tasks autonomously' including enriching records (twenty-docs-5, twenty-docs-56) plus workflow triggers on record changes (twenty-docs-3, twenty-docs-13). However, no evidence explicitly describes an agent reading email/meeting content and then updating deal stage or 'next steps' fields specifically — this exact pipeline-management workflow is not documented as a first-party example. Missing for 10: an explicit example/workflow showing an agent parsing email or meeting signals and writing deal-stage/next-steps updates via API or MCP, and any independent/hands-on confirmation this works end-to-end.

   • [claimed-docs] “MCP server | Yes | Yes | Yes”
   • [claimed-docs] “AI agents that work autonomously within your CRM — answering questions, enriching records, and executing multi-step tasks within your permis…”
   • [claimed-docs] “AI agents go beyond chat — they can execute multi-step tasks autonomously: Enrich records with data from external sources”
   • [claimed-docs] “Connect your Google Workspace or Microsoft 365 account. Emails and calendar events automatically appear on the relevant CRM records.”
   • [claimed-docs] “Connect your **Google Workspace** or **Microsoft 365** mailbox. Once connected: Emails are automatically linked to the matching Company and …”
   • [claimed-docs] “A developer-first API that adapts to your custom data model. Both GraphQL and REST endpoints, with auto-generated documentation per works…”
   • [claimed-docs] “Workflows can chain any combination of: Record operations... Send email... HTTP request... Code... Branches... Iterator... AI Agent... Delay…”
   • [claimed-docs] “Trigger workflows on record changes, schedules, manual actions, or incoming webhooks.”

   proof ↗⚑ flag

Deals

1. ops userRun multiple pipelines with different stage sets for different products or motions

   weight 2 · round to Twenty
   HubSpot—none0/10

   The evidence pack covers CRM data, API objects, MCP servers, workflows, reporting, and deduplication, but contains no mention of deal pipelines, multiple pipeline configuration, or custom stage sets per pipeline/product — the core capability the story asks about.

   ⚑ flag

   Twenty~partialclaimed5/10

   Twenty's docs show building blocks that could support multiple pipelines — unlimited custom objects with custom fields (twenty-docs-1, twenty-docs-33), kanban/table/calendar views with grouping and saved views (twenty-docs-34), and workflow triggers per object (twenty-docs-12) — but there is no explicit documentation of a native 'multiple pipelines with distinct stage sets' feature (e.g. a dedicated Opportunities/Deals module supporting several independently configured pipelines). missing for 10: explicit docs describing a pipeline/deal-stage object with support for multiple named pipelines each having its own stage set, and confirmation this works out-of-the-box rather than via manual custom-object/kanban configuration.

   • [claimed-docs] “Create custom objects, add custom fields with 20+ field types, and build relationships between any objects.”
   • [claimed-docs] “You can also create custom objects for anything your business needs — projects, support tickets, products, contracts, or anything else.”
   • [claimed-docs] “Switch between table views, kanban boards, and calendar views. Filter with AND/OR logic, sort by multiple fields, group records, and save cu…”
   • [claimed-docs] “Custom objects get the same first-class treatment as built-in ones — including API endpoints, views, permissions, and workflow triggers.”

   proof ↗⚑ flag

2. founderTrack deals on a visual pipeline with customizable stages and drag-and-drop updates

   weight 3 · round to Twenty
   HubSpot—none0/10

   The evidence pack contains extensive HubSpot CRM/API/reporting docs but never describes a visual deal pipeline, customizable stages, or drag-and-drop pipeline UI — only generic references to 'deals' as an object type in the MCP/CRM API docs. Since this is a well-known and applicable CRM capability but no supporting evidence is present in the pack, it must be scored as none rather than assumed.

   • [claimed-docs] “The HubSpot MCP server enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — conta…”
   • [claimed-docs] “enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — contacts, deals, engagements…”
   • [claimed-docs] “HubSpot's free CRM unifies all your customer data on one platform, with AI that makes it easy to understand. And unlike a free trial, it's 1…”

   proof ↗⚑ flag

   Twenty~partialcommunity7/10

   Docs confirm kanban board views alongside table/calendar views, customizable fields/objects for building deal stages, and workflows that can trigger on stage changes (e.g., Slack alert when 'a deal reaches a certain stage'), which together support a visual pipeline with configurable stages. However, no evidence explicitly confirms drag-and-drop interaction on the kanban board itself, and community feedback flags the product as 'pretty alpha-version' with unspecified issues. Missing for 10: explicit documentation/screenshot of drag-and-drop card movement between stages, and independent hands-on confirmation that pipeline stage customization works smoothly in practice.

   • [claimed-docs] “Switch between table views, kanban boards, and calendar views.”
   • [claimed-docs] “Switch between table views, kanban boards, and calendar views. Filter with AND/OR logic, sort by multiple fields, group records, and save cu…”
   • [claimed-docs] “Send Slack alerts when a deal reaches a certain stage * Auto-enrich new contacts with data from external APIs”
   • [claimed-docs] “Workflows can chain any combination of: Record operations... Send email... HTTP request... Code... Branches... Iterator... AI Agent... Delay…”
   • [community] “Twenty was almost what we needed. It was extremely easy to set up, the customization is very promising, and the REST API worked ok. The main…”

   proof ↗⚑ flag

Forecasting

1. founderForecast revenue from pipeline value, stage probabilities, and close dates

   weight 2 · round to Twenty
   HubSpot—none0/10

   The evidence pack shows generic sales reporting/dashboards (pre-built reports, custom dashboards, deal-related analytics) but nothing documents a dedicated pipeline forecasting feature that combines deal value, stage probability, and close dates into a revenue forecast. Without that specific capability being evidenced, this applicable CRM axis is unsupported.

   • [claimed-docs] “dozens of free pre-built sales reports that come standard with HubSpot’s sales reporting software”
   • [claimed-docs] “Choose from dozens of pre-built reports or customize your own”
   • [claimed-docs] “Build marketing, sales, revenue, and service analytics reports with filters and chart types to see exactly what's happening across your busi…”
   • [claimed-docs] “Monitor your team’s sales activities and unlock additional rep performance with deep coaching insights”

   proof ↗⚑ flag

   Twenty~partialclaimed5/10

   Twenty provides the building blocks—custom fields for deal value/stage/close date, dashboards with configurable charts, filters, aggregation, and grouping, and docs explicitly mention tracking 'pipeline health, team performance, revenue trends'—but there is no dedicated forecasting feature that models stage-probability-weighted revenue projections; a founder would need to manually configure fields and dashboard aggregations to approximate this. Missing for 10: a native weighted-pipeline forecast report, explicit stage-probability calculation logic, and any hands-on evidence of this being used for revenue forecasting.

   • [claimed-docs] “Build custom dashboards with real-time widgets. Track pipeline metrics, team performance, and business KPIs with configurable charts and …”
   • [claimed-docs] “You can configure: Chart type... Data source... Filters... Aggregation... Grouping”
   • [claimed-docs] “Dashboards give you real-time visibility into your business metrics — pipeline health, team performance, revenue trends”
   • [claimed-docs] “Create custom objects, add custom fields with 20+ field types, and build relationships between any objects.”
   • [claimed-docs] “Build custom dashboards with real-time widgets.”

   proof ↗⚑ flag

1. ai-native userChoose where my data is stored (region/residency)

   weight 2 · round to Twenty
   HubSpot—none0/10

   No evidence in the pack addresses data residency, regional data centers, or storage location options; all citations concern APIs, MCP, CRM features, imports, and general commentary. Missing for 10: any documentation of EU/US data hosting options, residency controls, or region selection for data storage.

   ⚑ flag

   Twenty~partialclaimed5/10

   Twenty doesn't offer an explicit region-selection setting, but self-hosting via Docker Compose lets privacy-conscious users run the entire stack on their own infrastructure, effectively letting them choose where data lives. This is an indirect residency control rather than a built-in region picker for the cloud offering. Missing for 10: explicit cloud data-residency/region selection UI, documented data center regions, compliance certifications (GDPR/SOC2) tied to specific regions.

   • [claimed-docs] “Run Twenty on your own infrastructure with a single Docker Compose command.”
   • [claimed-docs] “Privacy-conscious organizations — you need to self-host and own your data end to end. Regulatory, contractual, or just principle.”
   • [claimed-docs] “**Privacy-conscious organizations** — you need to self-host and own your data end to end.”
   • [claimed-docs] “This guide provides step-by-step instructions to install and configure the Twenty application using Docker Compose.”

   proof ↗⚑ flag

2. ai-native userPrevent my data from being used to train AI models

   weight 3 · round to Twenty
   HubSpot—none0/10

   No evidence pack item addresses AI training data usage, opt-out controls, or privacy policy language about AI model training; the evidence covers unrelated product features (CRM, MCP, reporting, workflows).

   ⚑ flag

   Twenty~partialclaimed4/10

   Twenty offers self-hosting so an organization can keep full control of its data ('own your data end to end'), which indirectly prevents third-party AI training on it, and permission-scoped AI agent access limits what data agents can see. However, there is no explicit statement, opt-out setting, or policy about excluding user data from AI model training on the cloud/hosted offering. missing for 10: explicit no-training data policy, cloud-hosted opt-out toggle, independent confirmation of data-handling practices for AI features.

   • [claimed-docs] “Privacy-conscious organizations — you need to self-host and own your data end to end. Regulatory, contractual, or just principle.”
   • [claimed-docs] “**Privacy-conscious organizations** — you need to self-host and own your data end to end.”
   • [claimed-docs] “Agents can only access objects and fields that the user (or role) has permission to view. Sensitive data stays protected even when AI is inv…”
   • [claimed-docs] “Run Twenty on your own infrastructure with a single Docker Compose command.”

   proof ↗⚑ flag

3. ai-native userControl data retention and deletion

   weight 2 · round to Twenty
   HubSpot—none0/10

   No evidence pack items address data retention policies, GDPR/CCPA deletion requests, data export controls, or configurable retention windows for HubSpot CRM/AI data — the story is applicable to a CRM/AI platform but unevidenced here.

   ⚑ flag

   Twenty~partialclaimed3/10

   Twenty offers data export ('Export your data anytime — no lock-in', up to 20,000 records/export) and self-hosting for full data ownership, which gives some control over data, but there is no explicit documentation of retention policies or deletion controls (e.g., record purge, right-to-be-forgotten, TTL settings) tailored to AI-native use. Missing for 10: explicit data retention policy settings, record/field deletion controls, GDPR-style erasure workflow, and any AI-specific retention/deletion controls.

   • [claimed-docs] “Export your data anytime — no lock-in.”
   • [claimed-docs] “Export your workspace data for backups, reporting, or migration. ... Export limits: Up to 20,000 records per export.”
   • [claimed-docs] “Privacy-conscious organizations — you need to self-host and own your data end to end. Regulatory, contractual, or just principle.”
   • [claimed-docs] “**Privacy-conscious organizations** — you need to self-host and own your data end to end.”

   proof ↗⚑ flag

4. ai-native userOpt out of telemetry and usage tracking

   weight 2 · round drawn
   HubSpot—none0/10

   No evidence in the pack addresses telemetry or usage-tracking opt-out settings for HubSpot; the docs cover CRM, API, MCP, and product features but nothing about privacy/telemetry controls.

   ⚑ flag

   Twenty—none0/10

   No evidence in the pack mentions telemetry, usage tracking, analytics collection, or an opt-out setting; self-hosting is mentioned but does not itself confirm control over telemetry data collection.

   ⚑ flag

Ai crm ops

1. ai-native userHave an agent compile an account brief — open deals, recent emails, and history — from CRM data via the API or MCP server before a meeting

   weight 2 · round drawn
   HubSpot~partialprobed6/10

   HubSpot's official MCP server explicitly grants read access to contacts, deals, and engagements, and its CRM APIs (contacts, custom objects, account info) plus connected-email logging supply the underlying data (open deals, emails, history) needed for such a brief; this directly supports an AI-native agent pulling account context via MCP or API before a meeting. However, there's no documented end-to-end 'account brief' workflow or example, no explicit deals/engagements API doc cited, and no independent/hands-on evidence the MCP server actually performs this compilation reliably. Missing for 10: a demonstrated brief-generation example, explicit deals/engagements endpoint docs, and third-party corroboration of MCP data-pull quality.

   • [claimed-docs] “The HubSpot MCP server enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — conta…”
   • [claimed-docs] “The HubSpot MCP server enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data”
   • [claimed-docs] “enables you to give any MCP-compatible AI tool or agent secure read and write access to your HubSpot CRM data — contacts, deals, engagements…”
   • [claimed-docs] “the contacts endpoints allow you to create and manage contact records in your HubSpot account, as well as sync contact data between HubSpot …”
   • [claimed-docs] “Connect an individual work email to HubSpot to send one-to-one emails from the CRM, log email replies to the CRM, send sequences emails, and…”
   • [claimed-docs] “Retrieve detailed information about the account associated with the HubSpot portal.”
   • [probe] “official MCP server documented at https://developers.hubspot.com/ai-tools/mcp”

   proof ↗⚑ flag

   Twenty~partialclaimed6/10

   Twenty provides the necessary building blocks — GraphQL/REST APIs and a documented MCP server (twenty-docs-32) for programmatic access, automatic linking of emails/calendar events to Company/Person records (twenty-docs-4, twenty-docs-52, twenty-docs-58), and AI agents that can answer questions, enrich records, and execute multi-step tasks within permissions (twenty-docs-5, twenty-docs-18, twenty-docs-56) — which together could let an agent assemble open deals, emails, and history into a brief. However, there is no concrete documentation of an actual 'account brief' workflow, no example of the MCP server's specific tools/resources for querying deals+emails+history in one call, and no hands-on evidence this composite use case has been built or tested. Missing for 10: a documented end-to-end example or MCP tool spec for compiling a pre-meeting brief, and independent/hands-on confirmation of the AI agent successfully synthesizing deals+emails+history.

   • [claimed-docs] “MCP server | Yes | Yes | Yes”
   • [claimed-docs] “AI agents that work autonomously within your CRM — answering questions, enriching records, and executing multi-step tasks within your permis…”
   • [claimed-docs] “Ask questions about your data in natural language. The AI chatbot can query your CRM records, summarize information, and help you find what …”
   • [claimed-docs] “AI agents go beyond chat — they can execute multi-step tasks autonomously: Enrich records with data from external sources”
   • [claimed-docs] “Connect your Google Workspace or Microsoft 365 account. Emails and calendar events automatically appear on the relevant CRM records.”
   • [claimed-docs] “Connect your **Google Workspace** or **Microsoft 365** mailbox. Once connected: Emails are automatically linked to the matching Company and …”
   • [claimed-docs] “Emails are automatically linked to the matching Company and People records”
   • [claimed-docs] “A developer-first API that adapts to your custom data model. Both GraphQL and REST endpoints, with auto-generated documentation per works…”

   proof ↗⚑ flag

Reporting

1. founderBuild reports and dashboards on pipeline, activity, and conversion without exporting to a spreadsheet

   weight 3 · round drawn
   HubSpot✓fullclaimed8/10

   HubSpot's reporting product explicitly supports building custom dashboards/reports on pipeline (sales reports, deal stages), activity (rep performance, coaching insights), and conversion, combining multiple reports per role, with natural-language report building via Breeze Assistant — all within the platform, explicitly framed against spreadsheets ('Your data deserves better than spreadsheets'). Missing for 10: independent/hands-on user corroboration of dashboard-building experience and no detail on conversion-funnel-specific reporting depth.

   • [claimed-docs] “combine multiple reports to help every team member view the data most relevant to their role”
   • [claimed-docs] “describe the report you need in plain language, and HubSpot will build it from your actual data.”
   • [claimed-docs] “Build custom reporting dashboards for every team”
   • [claimed-docs] “Choose from dozens of pre-built reports or customize your own”
   • [claimed-docs] “Build marketing, sales, revenue, and service analytics reports with filters and chart types to see exactly what's happening across your busi…”
   • [claimed-docs] “with Breeze Assistant, describe the report you need in plain language, and HubSpot will build it from your actual data.”
   • [claimed-docs] “Control who sees what with granular reporting permissions”
   • [claimed-docs] “Monitor your team’s sales activities and unlock additional rep performance with deep coaching insights”
   • [claimed-docs] “Your data deserves better than spreadsheets. HubSpot's free CRM unifies all your customer data on one platform... it's 100% free with no exp…”

   proof ↗⚑ flag

   Twenty✓fullcommunity8/10

   Twenty provides native custom dashboards with configurable chart types, filters, aggregation, and grouping over pipeline/activity data, with real-time widgets for pipeline health, team performance, and revenue trends — all built in-app without needing spreadsheet exports (twenty-docs-6, twenty-docs-20, twenty-docs-21, twenty-docs-57, twenty-docs-48). Data export is offered only as an optional side capability, not a requirement for reporting. Missing for 10: independent/hands-on validation of dashboard depth and reporting accuracy (only vendor docs cited, and community evidence [twenty-comm-1] notes general 'alpha-version' rough edges without specifically addressing dashboards).

   • [claimed-docs] “Build custom dashboards with real-time widgets. Track pipeline metrics, team performance, and business KPIs with configurable charts and …”
   • [claimed-docs] “You can configure: Chart type... Data source... Filters... Aggregation... Grouping”
   • [claimed-docs] “Dashboards are workspace-level — everyone on your team can see them. Arrange widgets in a grid layout and resize them.”
   • [claimed-docs] “Dashboards give you real-time visibility into your business metrics — pipeline health, team performance, revenue trends”
   • [claimed-docs] “Arrange widgets in a grid layout and resize them to build the view that works for your team.”
   • [claimed-docs] “Build custom dashboards with real-time widgets.”
   • [community] “Twenty was almost what we needed. It was extremely easy to set up, the customization is very promising, and the REST API worked ok. The main…”

   proof ↗⚑ flag