Jitsu vs Hightouch

Agent access

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

   weight 2 · round to Hightouch
   Jitsu✓fullprobed8/10

   A direct probe confirms llms.txt is live and served (HTTP 200) at https://docs.jitsu.com/llms.txt with a proper summary of Jitsu, and Jitsu also documents an MCP server for agent-driven pipeline management, showing genuine agent-oriented documentation surfaces. Missing for 10: no independent/community confirmation of an agent actually consuming llms.txt successfully, and no broader agent-facing docs index beyond the single llms.txt file.

   • [probe] “PROBE llms.txt: HTTP 200 at https://docs.jitsu.com/llms.txt # Jitsu > Jitsu is an open-source data integration platform and a self-hostable…”
   • [claimed-docs] “Jitsu runs an MCP server, so AI agents can manage your pipeline directly: create destinations, wire up streams, inspect Live Events, and edi…”
   • [probe] “official MCP server documented at https://jitsu.com/docs/mcp”
   • [probe] “PROBE runtime (recorded 2026-09-08): Jitsu's hosted MCP server ('Jitsu runs an MCP server, so AI agents can manage your pipeline directly: c…”

   proof ↗⚑ flag

   Hightouch✓fullprobed9/10

   A live llms.txt is confirmed at https://hightouch.com/llms.txt (HTTP 200), explicitly framing Hightouch as an 'Agentic Marketing Platform' and indexing its docs for agent consumption, corroborated by a runtime probe. This directly satisfies the story of pointing an agent at agent-oriented docs. Missing for 10: no docs.md/markdown-mirror endpoint (404) or OpenAPI spec discovery, which would round out full agent-native doc coverage.

   • [probe] “PROBE llms.txt: HTTP 200 at https://hightouch.com/llms.txt # Hightouch > Hightouch is an Agentic Marketing Platform powered by the industry…”
   • [probe] “PROBE runtime (recorded 2026-09-08): hightouch.com publishes an agent-oriented llms.txt on its main origin ('# Hightouch — Hightouch is an A…”
   • [probe] “PROBE docs-md: HTTP 404 at https://hightouch.com/docs.md”

   proof ↗⚑ flag

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

   weight 2 · round to Jitsu
   Jitsu✓fullprobed8/10

   Jitsu offers a jitsu-cli that scaffolds, tests, and deploys pipeline Functions from CI (jitsu-cli init/deploy), an HTTP API for backend/headless data sending, and an MCP server that explicitly documents authenticating via a personal API key (instead of browser flow) for CI/headless environments. Combined, these show clear support for headless/CI automation. missing for 10: no independent third-party CI pipeline example or case study showing jitsu-cli actually run inside a CI system end-to-end, and no evidence of a full OpenAPI/REST spec for scripting beyond the documented HTTP ingestion API.

   • [claimed-docs] “`jitsu-cli` is the command-line interface for Jitsu. It manages workspace configuration objects (destinations, streams, connections, …) and …”
   • [github] “`jitsu-cli init` scaffolds a TypeScript project with tests, and `jitsu-cli deploy` ships it to your workspace.”
   • [claimed-docs] “In CI and other headless environments the browser flow can't run. There you authenticate with a personal API key instead”
   • [claimed-docs] “Jitsu runs an MCP server, so AI agents can manage your pipeline directly: create destinations, wire up streams, inspect Live Events, and edi…”
   • [claimed-docs] “You can use HTTP API to send data to Jitsu. This is useful if you want to send data from your backend.”
   • [probe] “official CLI documented at https://jitsu.com/docs/jitsu-cli”
   • [probe] “PROBE runtime (recorded 2026-09-08): Jitsu's hosted MCP server ('Jitsu runs an MCP server, so AI agents can manage your pipeline directly: c…”

   proof ↗⚑ flag

   Hightouch~partialprobed6/10

   Hightouch exposes a REST API (auth-gated, verified live) for syncs/models/sources/destinations and supports Git Sync for programmatic resource management, both enabling scripted/CI-driven workflows. However, there's no dedicated CLI, no documented CI/CD pipeline examples, and syncs are primarily scheduled/triggered rather than designed for headless orchestration. missing for 10: official CLI tool, documented CI/CD integration examples or GitHub Actions templates, evidence of headless batch/automation runs outside the API, independent hands-on confirmation of CI usage.

   • [claimed-docs] “Hightouch exposes a REST API that lets users interact with resources like syncs, models, sources, and destinations.”
   • [claimed-docs] “Git Sync lets you manage workspace resources programmatically.”
   • [claimed-docs] “Authenticate each request by passing your API key as a bearer token in the `Authorization` header”
   • [probe] “PROBE runtime (recorded 2026-09-08): Hightouch's REST API is live and cleanly auth-gated — a keyless GET https://api.hightouch.com/api/v1/sy…”
   • [claimed-docs] “A sync runs on the schedule or trigger you set, then reports the outcome of each run so you can confirm it worked.”

   proof ↗⚑ flag

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

   weight 3 · round drawn
   Jitsu—none0/10

   Evidence shows Jitsu operates as an MCP *server* — 'Jitsu runs an MCP server, so AI agents can manage your pipeline directly' (jitsu-docs-19, jitsu-probe-rt-1) — which is the opposite role from what this story asks (Jitsu acting as an MCP *client* that plugs in and uses external MCP servers' tools). No evidence anywhere in the pack shows Jitsu consuming or connecting to third-party MCP servers to use their tools.

   • [claimed-docs] “Jitsu runs an MCP server, so AI agents can manage your pipeline directly: create destinations, wire up streams, inspect Live Events, and edi…”
   • [probe] “PROBE runtime (recorded 2026-09-08): Jitsu's hosted MCP server ('Jitsu runs an MCP server, so AI agents can manage your pipeline directly: c…”
   • [probe] “official MCP server documented at https://jitsu.com/docs/mcp”

   proof ↗⚑ flag

   Hightouch—none0/10

   All evidence describes Hightouch exposing its own MCP server so external AI assistants can call Hightouch's tools (server role), not Hightouch's built-in agent acting as an MCP client that can plug in and use other services' MCP servers/tools. No documentation or probe shows Hightouch's agent consuming third-party MCP servers.

   • [claimed-docs] “the Hightouch MCP server connects your AI assistant directly to your marketing stack so your team can build audiences, manage syncs, design …”
   • [claimed-docs] “Create audiences from warehouse data, build automated journeys, and use the built-in agent to explore segments and analyze campaign performa…”
   • [probe] “official MCP server documented at https://hightouch.com/docs/ai-integrations/mcp”
   • [probe] “PROBE runtime (recorded 2026-09-08): hightouch.com publishes an agent-oriented llms.txt on its main origin ('# Hightouch — Hightouch is an A…”

   proof ↗⚑ flag

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

   weight 3 · round drawn
   Jitsu✓fullprobed9/10

   Jitsu documents and runs an official MCP server (jitsu.com/docs/mcp) letting AI agents manage pipelines — create destinations, wire streams, inspect Live Events, edit Functions — with both browser OAuth and API-key auth flows for headless use. A runtime probe confirms the hosted MCP endpoint is live and enforces OAuth (401 with WWW-Authenticate), corroborating the documented capability. Missing for 10: independent third-party hands-on review of the MCP integration beyond vendor docs/probe.

   • [claimed-docs] “Jitsu runs an MCP server, so AI agents can manage your pipeline directly: create destinations, wire up streams, inspect Live Events, and edi…”
   • [claimed-docs] “In CI and other headless environments the browser flow can't run. There you authenticate with a personal API key instead”
   • [probe] “official MCP server documented at https://jitsu.com/docs/mcp”
   • [probe] “PROBE runtime (recorded 2026-09-08): Jitsu's hosted MCP server ('Jitsu runs an MCP server, so AI agents can manage your pipeline directly: c…”

   proof ↗⚑ flag

   Hightouch✓fullprobed9/10

   Hightouch documents an official MCP server that connects AI assistants directly to its marketing stack, supporting audience creation, sync management, journey design, and creative generation via natural language, with RBAC scoping. This is corroborated by both docs and a runtime probe confirming the MCP docs page and agent-oriented llms.txt. Missing for 10: independent hands-on third-party review of the MCP server in actual use.

   • [claimed-docs] “the Hightouch MCP server connects your AI assistant directly to your marketing stack so your team can build audiences, manage syncs, design …”
   • [claimed-docs] “"Create an audience of customers that added an item to their cart in the last day but haven't completed a transaction, then create a sync to…”
   • [claimed-docs] “Ad Studio Customers: Generate visual ad creatives from a prompt, refine them with follow-up instructions, and browse existing projects.”
   • [probe] “official MCP server documented at https://hightouch.com/docs/ai-integrations/mcp”
   • [probe] “PROBE runtime (recorded 2026-09-08): hightouch.com publishes an agent-oriented llms.txt on its main origin ('# Hightouch — Hightouch is an A…”

   proof ↗⚑ flag

5. ai-native userUse an official CLI

   weight 2 · round to Jitsu
   Jitsu✓fullprobed8/10

   Jitsu documents a dedicated jitsu-cli that manages workspace config objects and runs a Functions dev workflow, with init/deploy commands to scaffold and ship TypeScript projects, corroborated by both docs and GitHub README content. This is a genuine official CLI, not just a script wrapper, making it usable in agentic/CI workflows (e.g., alongside the MCP server for AI-agent pipeline management). Missing for 10: independent hands-on reviews of jitsu-cli specifically (community evidence discusses the platform broadly but not the CLI), and no detail on full command surface beyond init/deploy.

   • [claimed-docs] “`jitsu-cli` is the command-line interface for Jitsu. It manages workspace configuration objects (destinations, streams, connections, …) and …”
   • [claimed-docs] “jitsu-cli is the command-line interface for Jitsu. It manages workspace configuration objects (destinations, streams, connections, …) and ru…”
   • [github] “`jitsu-cli init` scaffolds a TypeScript project with tests, and `jitsu-cli deploy` ships it to your workspace.”
   • [github] “jitsu-cli init scaffolds a TypeScript project with tests, and jitsu-cli deploy ships it to your workspace.”
   • [github] “build and deploy them from your own repo with the Jitsu CLI — `jitsu-cli init` scaffolds a TypeScript project with tests, and `jitsu-cli dep…”
   • [probe] “official CLI documented at https://jitsu.com/docs/jitsu-cli”

   proof ↗⚑ flag

   Hightouch—none0/10

   The evidence pack covers Hightouch's REST API, Git Sync, and MCP server for AI assistants, but no official CLI tool is mentioned anywhere in the docs, pricing, or community evidence. Since Hightouch is a developer-facing data platform where a CLI would be a plausible and expected offering, its absence constitutes 'none' rather than 'na'.

   • [claimed-docs] “Hightouch exposes a REST API that lets users interact with resources like syncs, models, sources, and destinations.”
   • [claimed-docs] “Git Sync lets you manage workspace resources programmatically.”
   • [probe] “PROBE openapi: all candidate paths 404 (https://hightouch.com/openapi.json, https://hightouch.com/swagger.json, https://hightouch.com/api/op…”

   proof ↗⚑ flag

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

   weight 3 · round to Hightouch
   Jitsu~partialprobed6/10

   Jitsu documents an HTTP API for sending event data (jitsu-docs-2) and a first-party MCP server that lets AI agents manage pipelines, destinations, streams, and functions (jitsu-docs-19, jitsu-probe-3, jitsu-probe-rt-1 confirms it's live), plus a CLI for workspace config (jitsu-docs-13, jitsu-gh-1). However, there is no unified, documented OpenAPI/REST spec for full product control — probes for openapi/swagger endpoints all 404 (jitsu-probe-2), so 'driving the product' is split across a data-ingestion API, a separate MCP protocol, and a CLI rather than one cohesive public API. Missing for 10: a single documented OpenAPI/REST management API covering full product configuration, and independent hands-on confirmation of programmatic control beyond data ingestion.

   • [claimed-docs] “You can use HTTP API to send data to Jitsu. This is useful if you want to send data from your backend.”
   • [claimed-docs] “Jitsu runs an MCP server, so AI agents can manage your pipeline directly: create destinations, wire up streams, inspect Live Events, and edi…”
   • [probe] “official MCP server documented at https://jitsu.com/docs/mcp”
   • [probe] “PROBE runtime (recorded 2026-09-08): Jitsu's hosted MCP server ('Jitsu runs an MCP server, so AI agents can manage your pipeline directly: c…”
   • [claimed-docs] “`jitsu-cli` is the command-line interface for Jitsu. It manages workspace configuration objects (destinations, streams, connections, …) and …”
   • [github] “`jitsu-cli init` scaffolds a TypeScript project with tests, and `jitsu-cli deploy` ships it to your workspace.”
   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.jitsu.com/openapi.json, https://docs.jitsu.com/swagger.json, https://docs.jitsu.com/api…”

   proof ↗⚑ flag

   Hightouch✓fullprobed8/10

   Hightouch documents a REST API covering syncs, models, sources, and destinations with bearer-token authentication, and a runtime probe confirms the API is live and properly auth-gated. This gives AI-native users a real programmatic surface beyond the UI/MCP layer. Missing for 10: a discoverable OpenAPI/Swagger spec (probed and 404) and independent developer corroboration of API robustness.

   • [claimed-docs] “Hightouch exposes a REST API that lets users interact with resources like syncs, models, sources, and destinations.”
   • [claimed-docs] “Authenticate each request by passing your API key as a bearer token in the `Authorization` header”
   • [probe] “PROBE runtime (recorded 2026-09-08): Hightouch's REST API is live and cleanly auth-gated — a keyless GET https://api.hightouch.com/api/v1/sy…”
   • [probe] “PROBE openapi: all candidate paths 404 (https://hightouch.com/openapi.json, https://hightouch.com/swagger.json, https://hightouch.com/api/op…”

   proof ↗⚑ flag

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

   weight 2 · round drawn
   Jitsu~partialprobed4/10

   Jitsu supports personal API keys for headless/CI authentication and mentions 'named user API tokens with expiration' plus OAuth-protected MCP server access, suggesting some credential issuance controls, but there is no explicit documentation of scoped/least-privilege permission levels (e.g., read-only vs write, per-resource scopes) for these tokens or agent credentials. missing for 10: explicit scope/permission granularity for API tokens, documentation of least-privilege roles for agent-issued keys, independent confirmation of scoping enforcement.

   • [claimed-docs] “Audit log (SOC2-oriented) with account-activity alerts, plus named user API tokens with expiration.”
   • [claimed-docs] “In CI and other headless environments the browser flow can't run. There you authenticate with a personal API key instead”
   • [probe] “PROBE runtime (recorded 2026-09-08): Jitsu's hosted MCP server ('Jitsu runs an MCP server, so AI agents can manage your pipeline directly: c…”
   • [claimed-docs] “Jitsu runs an MCP server, so AI agents can manage your pipeline directly: create destinations, wire up streams, inspect Live Events, and edi…”

   proof ↗⚑ flag

   Hightouch~partialprobed4/10

   Hightouch supports bearer-token API keys for its REST API and permissions/RBAC/ownership boundaries, and MCP access is described as scoped to workspace RBAC, but there is no documented mechanism for issuing scoped, least-privilege, agent-specific credentials (e.g., per-agent API keys, granular scopes, or token minting for agents). missing for 10: dedicated scoped/least-privilege credential issuance for agents, documentation of API key scopes/permissions granularity, evidence of per-agent or short-lived token support.

   • [claimed-docs] “Authenticate each request by passing your API key as a bearer token in the `Authorization` header”
   • [claimed-docs] “Set up permissions, approvals, and ownership boundaries so each team has the right level of access.”
   • [claimed-docs] “the Hightouch MCP server connects your AI assistant directly to your marketing stack so your team can build audiences, manage syncs, design …”
   • [probe] “PROBE runtime (recorded 2026-09-08): hightouch.com publishes an agent-oriented llms.txt on its main origin ('# Hightouch — Hightouch is an A…”
   • [probe] “PROBE runtime (recorded 2026-09-08): Hightouch's REST API is live and cleanly auth-gated — a keyless GET https://api.hightouch.com/api/v1/sy…”

   proof ↗⚑ flag

8. ai-native userBuild against official SDKs

   weight 2 · round to Hightouch
   Jitsu~partialprobed6/10

   Jitsu ships an official JS/Segment-compatible SDK (@jitsu/js, confirmed installable via npm in jitsu-probe-rt-2), an HTTP API, and a jitsu-cli for functions/config, which covers 'official SDK' for AI-native integration builders. However, coverage is thin for broader agentic/AI-native SDK usage: no dedicated server-side SDKs beyond HTTP API, no OpenAPI spec (404s in jitsu-probe-2), and community feedback notes higher implementation friction versus competitors like Segment (jitsu-comm-6, jitsu-comm-7). missing for 10: published OpenAPI/REST SDK spec, broader multi-language official SDKs beyond JS/HTTP, independent hands-on validation of SDK ease-of-use matching docs claims.

   • [claimed-docs] “Jitsu javascript integration is 100% compatible with Segment API.”
   • [claimed-docs] “You can use HTTP API to send data to Jitsu. This is useful if you want to send data from your backend.”
   • [claimed-docs] “`jitsu-cli` is the command-line interface for Jitsu. It manages workspace configuration objects (destinations, streams, connections, …) and …”
   • [github] “`jitsu-cli init` scaffolds a TypeScript project with tests, and `jitsu-cli deploy` ships it to your workspace.”
   • [probe] “official CLI documented at https://jitsu.com/docs/jitsu-cli”
   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.jitsu.com/openapi.json, https://docs.jitsu.com/swagger.json, https://docs.jitsu.com/api…”
   • [probe] “PROBE runtime (recorded 2026-09-08): Jitsu is really self-hostable from source — a keyless shallow clone of jitsucom/jitsu landed the MIT LI…”
   • [community] “You've got most major ads sources that I care about, but it seems that there is a higher bar to implementation. Segment lets me just plug in…”
   • [community] “Functions are an underrated and highly powerful feature of Segment... It doesn't seem present here and that would hold a customer such as my…”

   proof ↗⚑ flag

   Hightouch✓fullprobed7/10

   Hightouch documents official SDKs for browser, iOS, Android, Node.js plus an HTTP API for event ingestion, and a separate authenticated REST API for managing syncs/models/sources/destinations, giving AI-native builders multiple official, documented integration surfaces. Missing for 10: a publicly discoverable OpenAPI/swagger spec (probe shows 404s across candidate paths) and independent/hands-on developer corroboration of SDK reliability beyond vendor docs.

   • [claimed-docs] “Hightouch provides SDKs for browser, iOS, Android, Node.js, and other platforms, plus an HTTP API for server-side ingestion”
   • [claimed-docs] “Hightouch provides SDKs for [browser](/docs/events/sdks/browser), [iOS](/docs/events/sdks/ios), [Android](/docs/events/sdks/android), [Node.…”
   • [claimed-docs] “Hightouch exposes a REST API that lets users interact with resources like syncs, models, sources, and destinations.”
   • [claimed-docs] “Authenticate each request by passing your API key as a bearer token in the `Authorization` header”
   • [claimed-docs] “You instrument your app, site, or backend with a Hightouch SDK, the HTTP API, or a streaming source like Kafka or Google Pub/Sub.”
   • [probe] “PROBE openapi: all candidate paths 404 (https://hightouch.com/openapi.json, https://hightouch.com/swagger.json, https://hightouch.com/api/op…”
   • [probe] “PROBE runtime (recorded 2026-09-08): Hightouch's REST API is live and cleanly auth-gated — a keyless GET https://api.hightouch.com/api/v1/sy…”

   proof ↗⚑ flag

9. ai-native userSubscribe to events via webhooks

   weight 2 · round drawn
   Jitsu—none0/10

   Evidence shows Jitsu can *receive* data via webhook (e.g., a Segment webhook destination sending data into Jitsu) and can deliver events to various destinations, but there is no evidence of an outbound webhook subscription mechanism that lets an AI-native consumer subscribe to Jitsu's own event stream. The MCP server evidence covers pipeline management, not webhook-based event subscription.

   • [claimed-docs] “Set up a webhook destination in Segment and send data to Jitsu and get it into your DB in real-time.”
   • [claimed-docs] “Tired of waiting 12 hours until your data is loaded to your warehouse? Set up a webhook destination in Segment and send data to Jitsu and ge…”
   • [github] “Delivery to destinations — ClickHouse, BigQuery, Snowflake, Redshift, Postgres, S3, GCS, and dozens of SaaS tools, streamed or micro-batched…”
   • [claimed-docs] “Jitsu runs an MCP server, so AI agents can manage your pipeline directly: create destinations, wire up streams, inspect Live Events, and edi…”

   proof ↗⚑ flag

   Hightouch—none0/10

   Evidence shows sync alerts/notifications (docs-28) and a REST API (docs-13, docs-33) but no documentation of a webhook subscription mechanism for events; Hightouch Events (docs-5/6/7) is about ingesting customer behavior data, not emitting webhooks for system/sync events. No evidence of an outbound webhook API or subscription endpoint for AI agents to consume.

   • [claimed-docs] “Get notified when a run fails [Configure sync alerts]”
   • [claimed-docs] “Hightouch exposes a REST API that lets users interact with resources like syncs, models, sources, and destinations.”
   • [claimed-docs] “Hightouch Events collects customer behavior from websites, mobile apps, and backend systems, then stores that data in your warehouse.”

   proof ↗⚑ flag

Agentic features

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

   weight 2 · round to Hightouch
   Jitsu—none0/10

   Jitsu's AI-related evidence is limited to an MCP server that lets external AI agents configure the pipeline (create destinations, streams, edit functions) — this is agentic control of infrastructure, not the product generating insights or suggestions from the data itself. No mention of built-in generative analytics, natural-language querying, anomaly detection, or AI-derived recommendations surfaced to the user inside Jitsu.

   • [claimed-docs] “Jitsu runs an MCP server, so AI agents can manage your pipeline directly: create destinations, wire up streams, inspect Live Events, and edi…”
   • [claimed-docs] “In CI and other headless environments the browser flow can't run. There you authenticate with a personal API key instead”
   • [probe] “PROBE runtime (recorded 2026-09-08): Jitsu's hosted MCP server ('Jitsu runs an MCP server, so AI agents can manage your pipeline directly: c…”
   • [claimed-docs] “Profile Builder generates profiles based on the `traits` object in `identify` events.”
   • [claimed-docs] “Profiles are customer records stored in your warehouse, based on the events data you send to Jitsu.”

   proof ↗⚑ flag

   Hightouch~partialclaimed6/10

   Hightouch documents a 'built-in agent' to explore segments and analyze campaign performance in natural language (hightouch-docs-19) and an AI Decisioning feature that uses reinforcement learning to recommend messages/channel/timing (hightouch-docs-14), which are AI-generated insights/suggestions surfaced in-product. However, most evidence emphasizes agentic actions (building audiences, syncs, generating ad creatives) via MCP rather than analytical insights/recommendations presented directly in the UI, and there's no independent/hands-on corroboration of the insights quality. missing for 10: independent verification of the built-in agent's insight/recommendation quality, more detail on how insights are surfaced in-product (dashboards, reports) beyond brief doc mentions, and hands-on evidence of AI Decisioning's real-world recommendation accuracy.

   • [claimed-docs] “AI Decisioning uses reinforcement learning to recommend the best message, channel, and timing for each customer, then measures what worked.”
   • [claimed-docs] “Create audiences from warehouse data, build automated journeys, and use the built-in agent to explore segments and analyze campaign performa…”
   • [claimed-docs] “the Hightouch MCP server connects your AI assistant directly to your marketing stack so your team can build audiences, manage syncs, design …”
   • [claimed-docs] “"Create an audience of customers that added an item to their cart in the last day but haven't completed a transaction, then create a sync to…”

   proof ↗⚑ flag

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

   weight 2 · round to Hightouch
   Jitsu~partialclaimed6/10

   Jitsu ships several mechanisms that run autonomously once configured: Functions process every incoming event without manual intervention (jitsu-docs-3, jitsu-docs-25, jitsu-docs-31), Connector syncs pull data on a schedule via Kubernetes CronJobs (jitsu-docs-41, jitsu-gh-3), and a dead-letter queue with a reprocessing worker retries failed deliveries automatically (jitsu-docs-16). These qualify as background automations that run without ongoing human action. Missing for 10: explicit scheduling/trigger configuration UI for non-self-hosted users, independent hands-on confirmation that these automations reliably run unattended over time, and clarity on how sync frequency/cadence is user-configurable.

   • [claimed-docs] “Functions are designed to process data in Jitsu before it is sent to the destination... Filter... Transform... Enrich”
   • [claimed-docs] “Modify, filter or augment events before it is stored in your warehouse with Jitsu Functions.”
   • [claimed-docs] “Filter, transform, and enrich events with TypeScript functions before they land.”
   • [claimed-docs] “A feature-complete Jitsu deployment requires a Kubernetes cluster: functions and profile builders run in dedicated function-server deploymen…”
   • [github] “Connector syncs — pull data *into* your warehouse from third-party sources (Airbyte-compatible connectors).”
   • [claimed-docs] “Dead-letter queue with a reprocessing worker: failed events can be replayed instead of lost.”

   proof ↗⚑ flag

   Hightouch✓fullclaimed7/10

   Hightouch supports scheduled/triggered syncs and automated journeys that run continuously without manual intervention, plus AI Decisioning that uses reinforcement learning to autonomously choose message/channel/timing and measure outcomes, and CDC-based background data propagation — all core 'set it and forget it' automations. missing for 10: no evidence of autonomous multi-step agentic workflows beyond marketing ops (e.g., self-initiated remediation, cross-tool agent chains) and no independent hands-on confirmation of long-running autonomous behavior.

   • [claimed-docs] “A sync runs on the schedule or trigger you set, then reports the outcome of each run so you can confirm it worked.”
   • [claimed-docs] “AI Decisioning uses reinforcement learning to recommend the best message, channel, and timing for each customer, then measures what worked.”
   • [claimed-docs] “Create audiences from warehouse data, build automated journeys, and use the built-in agent to explore segments and analyze campaign performa…”
   • [claimed-docs] “Sync modeled data to sales, support, analytics, and internal tools on a schedule or when data changes.”
   • [claimed-docs] “Instead of sending the full query result on every run, CDC compares the current results to the previous run and only sends new, changed, and…”
   • [claimed-docs] “For higher performance at scale, you can enable the Lightning sync engine to compute CDC directly in the warehouse.”
   • [claimed-docs] “Get notified when a run fails [Configure sync alerts]”

   proof ↗⚑ flag

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

   weight 3 · round to Hightouch
   Jitsu—none0/10

   Jitsu documents an MCP server that lets external AI agents control the pipeline (create destinations, edit functions, etc.), but this is the reverse of the story — it makes Jitsu controllable BY agents, not a built-in assistant users delegate tasks to within Jitsu's own UI. No evidence of an in-product AI assistant/copilot feature exists in the pack.

   • [claimed-docs] “Jitsu runs an MCP server, so AI agents can manage your pipeline directly: create destinations, wire up streams, inspect Live Events, and edi…”
   • [probe] “PROBE runtime (recorded 2026-09-08): Jitsu's hosted MCP server ('Jitsu runs an MCP server, so AI agents can manage your pipeline directly: c…”

   proof ↗⚑ flag

   Hightouch~partialclaimed5/10

   Docs explicitly mention a 'built-in agent' for exploring segments and analyzing campaign performance in natural language (hightouch-docs-19), and AI Decisioning uses reinforcement learning as an automated in-product agent (hightouch-docs-14). However, most of the detailed AI-assistant documentation (hightouch-docs-10/11/12/31, probe-4) actually describes the MCP server letting external AI assistants control Hightouch, not a dedicated in-app assistant UI with its own docs page. Missing for 10: a dedicated feature page describing the built-in assistant's UI/workflow, independent hands-on reports of using it, and detail on its scope/limitations.

   • [claimed-docs] “Create audiences from warehouse data, build automated journeys, and use the built-in agent to explore segments and analyze campaign performa…”
   • [claimed-docs] “AI Decisioning uses reinforcement learning to recommend the best message, channel, and timing for each customer, then measures what worked.”
   • [claimed-docs] “the Hightouch MCP server connects your AI assistant directly to your marketing stack so your team can build audiences, manage syncs, design …”

   proof ↗⚑ flag

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

   weight 2 · round drawn
   Jitsu✓fullprobed8/10

   Jitsu documents and runtime-verifies a first-party MCP server that lets AI agents manage the pipeline directly—creating destinations, wiring streams, inspecting Live Events, and editing Functions—which is the mechanism for natural-language/agentic operation, plus a documented headless auth path for CI use. Missing for 10: independent/hands-on evidence of actual natural-language command sessions in production and more detail on the breadth of commands the MCP server understands beyond the listed pipeline actions.

   • [claimed-docs] “Jitsu runs an MCP server, so AI agents can manage your pipeline directly: create destinations, wire up streams, inspect Live Events, and edi…”
   • [claimed-docs] “In CI and other headless environments the browser flow can't run. There you authenticate with a personal API key instead”
   • [probe] “official MCP server documented at https://jitsu.com/docs/mcp”
   • [probe] “PROBE runtime (recorded 2026-09-08): Jitsu's hosted MCP server ('Jitsu runs an MCP server, so AI agents can manage your pipeline directly: c…”

   proof ↗⚑ flag

   Hightouch✓fullprobed8/10

   Hightouch documents a first-party MCP server enabling natural-language commands to build audiences, manage syncs, design journeys, and generate creatives (e.g. 'Create an audience... then create a sync to Braze'), plus a built-in in-product agent for exploring segments/analyzing campaigns in natural language. This is corroborated by an llms.txt describing Hightouch as an 'Agentic Marketing Platform' and runtime probes confirming the MCP docs and API exist. Missing for 10: independent/hands-on user reports specifically validating the MCP/agent natural-language workflows (only vendor docs and generic HN commentary on other features exist).

   • [claimed-docs] “the Hightouch MCP server connects your AI assistant directly to your marketing stack so your team can build audiences, manage syncs, design …”
   • [claimed-docs] “"Create an audience of customers that added an item to their cart in the last day but haven't completed a transaction, then create a sync to…”
   • [claimed-docs] “Create audiences from warehouse data, build automated journeys, and use the built-in agent to explore segments and analyze campaign performa…”
   • [probe] “official MCP server documented at https://hightouch.com/docs/ai-integrations/mcp”
   • [probe] “PROBE runtime (recorded 2026-09-08): hightouch.com publishes an agent-oriented llms.txt on its main origin ('# Hightouch — Hightouch is an A…”

   proof ↗⚑ flag

Api quality

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

   weight 2 · round drawn
   Jitsu—none0/10

   No evidence of an interactive API reference with runnable examples; only static HTTP API docs mentioning sending data via HTTP, and probes for OpenAPI/Swagger specs at Jitsu's docs domain all returned 404s.

   • [claimed-docs] “You can use HTTP API to send data to Jitsu. This is useful if you want to send data from your backend.”
   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.jitsu.com/openapi.json, https://docs.jitsu.com/swagger.json, https://docs.jitsu.com/api…”

   proof ↗⚑ flag

   Hightouch—none0/10

   Hightouch documents a REST API with auth guidance (hightouch-docs-13, hightouch-docs-33) but there is no evidence of an interactive API reference with runnable examples — probes for openapi.json/swagger.json all returned 404 across every candidate path, and no docs mention a try-it console, Postman collection, or embedded API explorer.

   • [claimed-docs] “Hightouch exposes a REST API that lets users interact with resources like syncs, models, sources, and destinations.”
   • [claimed-docs] “Authenticate each request by passing your API key as a bearer token in the `Authorization` header”
   • [probe] “PROBE openapi: all candidate paths 404 (https://hightouch.com/openapi.json, https://hightouch.com/swagger.json, https://hightouch.com/api/op…”

   proof ↗⚑ flag

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

   weight 2 · round drawn
   Jitsu—none0/10

   A direct probe for OpenAPI/Swagger spec files at all standard locations returned 404s, and no evidence pack item documents a downloadable machine-readable API spec; only an llms.txt was found which is not an API spec.

   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.jitsu.com/openapi.json, https://docs.jitsu.com/swagger.json, https://docs.jitsu.com/api…”

   proof ↗⚑ flag

   Hightouch—none0/10

   Hightouch has a REST API (hightouch-docs-13, hightouch-probe-rt-1) but explicit probes for an OpenAPI/Swagger spec at all standard paths returned 404 (hightouch-probe-3), and no docs page offers a downloadable machine-readable spec.

   • [probe] “PROBE openapi: all candidate paths 404 (https://hightouch.com/openapi.json, https://hightouch.com/swagger.json, https://hightouch.com/api/op…”
   • [claimed-docs] “Hightouch exposes a REST API that lets users interact with resources like syncs, models, sources, and destinations.”

   proof ↗⚑ flag

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

   weight 1 · round to Jitsu
   Jitsu~partialclaimed5/10

   Jitsu offers a functions debugger that runs on sample data (jitsu-docs-38) and a zero-config development Helm chart for spinning up the full architecture on Minikube (jitsu-docs-15), both of which let a user test pipeline logic without touching production data. However, there's no explicit documented 'sandbox environment' or staging/prod data-isolation feature, and no AI-agent-specific guidance on using these dev tools safely. Missing for 10: explicit sandbox/staging environment concept, documented separation of test vs production data stores, and AI-agent-oriented sandbox testing workflow.

   • [claimed-docs] “Jitsu comes with a functions debugger/editor that allows to run function on a sample data”
   • [claimed-docs] “Development Helm chart: deploy the full architecture to Minikube with zero configuration.”
   • [claimed-docs] “A feature-complete Jitsu deployment requires a Kubernetes cluster: functions and profile builders run in dedicated function-server deploymen…”

   proof ↗⚑ flag

   Hightouch—none0/10

   No evidence in the pack of a sandbox/staging environment or test mode that isolates AI-native testing from production data; docs cover syncs, models, identity resolution, MCP, and API auth, but nothing about a non-production sandbox for testing.

   ⚑ flag

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

   weight 2 · round drawn
   Jitsu—none0/10

   No evidence of API versioning scheme (e.g., v1/v2 endpoints) or any documented deprecation policy for Jitsu's HTTP API, MCP server, or CLI; the openapi.json probe returned 404s and no changelog entry references versioning/deprecation commitments. Missing for 10: documented API version scheme, explicit deprecation policy/notice process, evidence of backward-compatibility guarantees.

   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.jitsu.com/openapi.json, https://docs.jitsu.com/swagger.json, https://docs.jitsu.com/api…”
   • [claimed-docs] “You can use HTTP API to send data to Jitsu. This is useful if you want to send data from your backend.”
   • [claimed-docs] “Jitsu runs an MCP server, so AI agents can manage your pipeline directly: create destinations, wire up streams, inspect Live Events, and edi…”

   proof ↗⚑ flag

   Hightouch—none0/10

   Evidence confirms Hightouch has a REST API (bearer-token auth, resources like syncs/models) but nothing in the docs pack mentions API versioning scheme (e.g., v1/v2) or any documented deprecation policy; OpenAPI spec discovery even failed (404s). missing for 10: any mention of API version numbers, changelog, or deprecation/sunset policy for the REST API or MCP server.

   • [claimed-docs] “Hightouch exposes a REST API that lets users interact with resources like syncs, models, sources, and destinations.”
   • [claimed-docs] “Authenticate each request by passing your API key as a bearer token in the `Authorization` header”
   • [probe] “PROBE openapi: all candidate paths 404 (https://hightouch.com/openapi.json, https://hightouch.com/swagger.json, https://hightouch.com/api/op…”
   • [probe] “PROBE runtime (recorded 2026-09-08): Hightouch's REST API is live and cleanly auth-gated — a keyless GET https://api.hightouch.com/api/v1/sy…”

   proof ↗⚑ flag

Agent audiences

1. ai-native userAn agent can query customer data and create or activate audiences end to end through documented APIs or MCP — no dashboard in the loop

   weight 3 · round to Hightouch
   Jitsu~partialprobed4/10

   Jitsu documents a real MCP server (jitsu-docs-19, confirmed live via jitsu-probe-rt-1) that lets agents manage pipeline objects — destinations, streams, functions, live events — and Profile Builder/warehouse query capabilities (jitsu-docs-11/12/34) exist for building customer profiles from event data. However, there is no documented API/MCP action specifically for creating or activating 'audiences' (segments for marketing activation) — the MCP scope is pipeline configuration, not audience management, and profile generation is not shown as agent-callable end to end. Missing for 10: explicit audience/segment creation endpoint, an 'activate audience' API or MCP tool, and evidence an agent can query already-built customer profiles/audiences via API without dashboard.

   • [claimed-docs] “Jitsu runs an MCP server, so AI agents can manage your pipeline directly: create destinations, wire up streams, inspect Live Events, and edi…”
   • [probe] “PROBE runtime (recorded 2026-09-08): Jitsu's hosted MCP server ('Jitsu runs an MCP server, so AI agents can manage your pipeline directly: c…”
   • [claimed-docs] “Profile Builder generates profiles based on the `traits` object in `identify` events.”
   • [claimed-docs] “`ctx.getWarehouse` - warehouse API to query your data warehouses”
   • [claimed-docs] “Profiles are customer records stored in your warehouse, based on the events data you send to Jitsu.”

   proof ↗⚑ flag

   Hightouch✓fullprobed7/10

   Hightouch documents an official MCP server that lets an AI assistant directly create audiences, build syncs to activate them (e.g., to Braze), and even generate creatives — explicitly in natural language with no dashboard step described (hightouch-docs-10/11/12/31, hightouch-probe-4, hightouch-probe-rt-2). This is reinforced by a documented, auth-gated REST API covering syncs/models/sources/destinations (hightouch-docs-13, hightouch-probe-rt-1) for programmatic control beyond MCP. missing for 10: independent/hands-on verification that MCP-driven audience creation works end-to-end without any dashboard dependency, and clarity on whether initial schema/model setup (docs-1/24/35 imply UI-based schema config precedes audience building) can be fully done via MCP/API alone.

   • [claimed-docs] “the Hightouch MCP server connects your AI assistant directly to your marketing stack so your team can build audiences, manage syncs, design …”
   • [claimed-docs] “"Create an audience of customers that added an item to their cart in the last day but haven't completed a transaction, then create a sync to…”
   • [claimed-docs] “Ad Studio Customers: Generate visual ad creatives from a prompt, refine them with follow-up instructions, and browse existing projects.”
   • [claimed-docs] “Hightouch exposes a REST API that lets users interact with resources like syncs, models, sources, and destinations.”
   • [probe] “official MCP server documented at https://hightouch.com/docs/ai-integrations/mcp”
   • [probe] “PROBE runtime (recorded 2026-09-08): Hightouch's REST API is live and cleanly auth-gated — a keyless GET https://api.hightouch.com/api/v1/sy…”
   • [probe] “PROBE runtime (recorded 2026-09-08): hightouch.com publishes an agent-oriented llms.txt on its main origin ('# Hightouch — Hightouch is an A…”

   proof ↗⚑ flag

Agent pipeline

1. ai-native userAn agent can manage the pipeline itself — create sources and destinations, wire streams, inspect deliveries — through a documented API or MCP server

   weight 3 · round to Jitsu
   Jitsu✓fullprobed8/10

   Jitsu explicitly documents a hosted MCP server letting AI agents create destinations, wire streams, inspect Live Events, and edit Functions, with API-key auth for headless/CI use — and a runtime probe confirms the MCP endpoint is live and enforces OAuth/API-key auth as documented. This directly matches the story's ask for agent-driven pipeline management via MCP. Missing for 10: independent hands-on validation of an agent actually performing pipeline edits end-to-end, and a general REST/OpenAPI spec (probe found no OpenAPI endpoint) that would complement the MCP path.

   • [claimed-docs] “Jitsu runs an MCP server, so AI agents can manage your pipeline directly: create destinations, wire up streams, inspect Live Events, and edi…”
   • [claimed-docs] “In CI and other headless environments the browser flow can't run. There you authenticate with a personal API key instead”
   • [probe] “official MCP server documented at https://jitsu.com/docs/mcp”
   • [probe] “PROBE runtime (recorded 2026-09-08): Jitsu's hosted MCP server ('Jitsu runs an MCP server, so AI agents can manage your pipeline directly: c…”

   proof ↗⚑ flag

   Hightouch✓fullprobed7/10

   Hightouch documents both an official MCP server (build audiences, manage syncs, design journeys, generate creatives, analyze performance) and a REST API explicitly covering syncs, models, sources, and destinations, plus a live authenticated API endpoint confirmed via probe. This directly satisfies the story of an agent managing the pipeline via API/MCP, including inspecting sync run outcomes/debugging. Missing for 10: explicit MCP-level source/destination creation (MCP docs emphasize audiences/syncs/journeys, not raw source/destination wiring), no discoverable OpenAPI spec, and no independent/hands-on evidence of an agent actually performing end-to-end pipeline management.

   • [claimed-docs] “the Hightouch MCP server connects your AI assistant directly to your marketing stack so your team can build audiences, manage syncs, design …”
   • [claimed-docs] “"Create an audience of customers that added an item to their cart in the last day but haven't completed a transaction, then create a sync to…”
   • [claimed-docs] “Hightouch exposes a REST API that lets users interact with resources like syncs, models, sources, and destinations.”
   • [claimed-docs] “A sync connects one model to one destination object, list, or table, and describes how the model's rows map into that destination.”
   • [claimed-docs] “Check a run or debug rejected rows | [Debug a sync run](/docs/syncs/debugger)”
   • [probe] “official MCP server documented at https://hightouch.com/docs/ai-integrations/mcp”
   • [probe] “PROBE runtime (recorded 2026-09-08): Hightouch's REST API is live and cleanly auth-gated — a keyless GET https://api.hightouch.com/api/v1/sy…”
   • [probe] “PROBE runtime (recorded 2026-09-08): hightouch.com publishes an agent-oriented llms.txt on its main origin ('# Hightouch — Hightouch is an A…”

   proof ↗⚑ flag

Activation

1. marketerAudiences sync to ad platforms and engagement tools continuously, with membership entering and exiting in near-real-time

   weight 2 · round to Hightouch
   Jitsu—none0/10

   Jitsu documents destinations (incl. SaaS ad tools), streaming/micro-batched delivery, and a Profile Builder that computes customer profiles from events/traits — but there is no evidence of an audience/segment-membership construct that continuously syncs entering/exiting members to ad platforms specifically. A community reviewer even notes ad-platform destinations require more manual work than Segment's plug-and-play audience sync (jitsu-comm-6).

   • [github] “Delivery to destinations — ClickHouse, BigQuery, Snowflake, Redshift, Postgres, S3, GCS, and dozens of SaaS tools, streamed or micro-batched…”
   • [claimed-docs] “Send data to one of those destinations with Jitsu. Learn more about destination on docs website”
   • [claimed-docs] “Profile Builder generates profiles based on the `traits` object in `identify` events.”
   • [claimed-docs] “Profile Builder generates profiles based on the traits object in identify events. You can also define custom logic for profile generation us…”
   • [claimed-docs] “Profiles are customer records stored in your warehouse, based on the events data you send to Jitsu.”
   • [community] “You've got most major ads sources that I care about, but it seems that there is a higher bar to implementation. Segment lets me just plug in…”

   proof ↗⚑ flag

   Hightouch~partialclaimed6/10

   Hightouch supports syncing audiences to ad platforms/engagement tools on a schedule or trigger with CDC to only send new/changed/removed rows, which supports near-real-time membership changes, but syncs are still schedule/trigger-based batch jobs rather than a documented continuous streaming/real-time audience membership pipeline. missing for 10: explicit documentation of true real-time/streaming sync latency (e.g. sub-minute or event-triggered entry/exit), independent evidence of sync frequency/latency in production, and confirmation that ad platform destinations support such near-real-time updates rather than periodic batch syncs.

   • [claimed-docs] “A sync runs on the schedule or trigger you set, then reports the outcome of each run so you can confirm it worked.”
   • [claimed-docs] “Instead of sending the full query result on every run, CDC compares the current results to the previous run and only sends new, changed, and…”
   • [claimed-docs] “For higher performance at scale, you can enable the Lightning sync engine to compute CDC directly in the warehouse.”
   • [claimed-docs] “A sync connects one model to one destination object, list, or table, and describes how the model's rows map into that destination.”
   • [claimed-docs] “Sync modeled data to sales, support, analytics, and internal tools on a schedule or when data changes.”

   proof ↗⚑ flag

Audiences

1. marketerBuild audiences from traits and behavior in a visual builder — no SQL required — and see estimated size before activating

   weight 3 · round to Hightouch
   Jitsu—none0/10

   Jitsu's evidence shows a 'Profile Builder' that generates customer profiles from traits/events, but this is developer-driven (JS functions) and warehouse/SQL-oriented, not a marketer-facing visual audience builder with no-SQL segment creation or size-before-activation preview. Docs even show raw SQL as the querying mechanism (jitsu-docs-43), the opposite of a no-SQL builder. No evidence of a visual campaign/audience UI or size estimation feature anywhere in the pack.

   • [claimed-docs] “Profile Builder generates profiles based on the `traits` object in `identify` events.”
   • [claimed-docs] “You can also define custom logic for profile generation using a JavaScript function, allowing you to leverage up to a year’s worth of user e…”
   • [claimed-docs] “Profiles are customer records stored in your warehouse, based on the events data you send to Jitsu.”
   • [claimed-docs] “select * from events where context_page_path = '/signup' and timestamp > now() - interval '1 day'”

   proof ↗⚑ flag

   Hightouch~partialclaimed6/10

   Docs confirm a no-code visual builder for audiences (Customer Studio) built on traits/behavior data, explicitly requiring no SQL from marketers, and models can also be built visually per hightouch-docs-23. However, none of the evidence mentions an audience size estimate/preview before activation. Missing for 10: explicit documentation or screenshot of estimated audience size display prior to activating a sync, and independent/hands-on confirmation of the visual builder UX.

   • [claimed-docs] “marketers can build audiences and journeys in a no-code UI. Audiences cannot be created until this step is complete.”
   • [claimed-docs] “Once the schema is configured, marketers can build audiences and journeys in a no-code UI.”
   • [claimed-docs] “Once the schema is configured, marketers can build [audiences](/docs/customer-studio/usage) and [journeys](/docs/customer-studio/journeys) i…”
   • [claimed-docs] “Create audiences from warehouse data, build automated journeys, and use the built-in agent to explore segments and analyze campaign performa…”
   • [claimed-docs] “Models — Define reusable datasets with SQL, dbt, or a visual builder”

   proof ↗⚑ flag

Traits

1. marketerComputed traits and predictive scores (LTV, churn or purchase propensity) are calculated on profiles and usable in targeting

   weight 2 · round to Hightouch
   Jitsu~partialclaimed3/10

   Jitsu's Profile Builder lets you compute custom traits via a JavaScript function against up to a year of event history, which covers 'computed traits' in principle, but there is no evidence of any built-in predictive scoring (LTV, churn, purchase propensity) or ML capability — this would require the marketer to hand-roll such logic themselves, and no example or feature is documented for it. missing for 10: built-in predictive/ML scoring models, documented LTV/churn/propensity outputs, and evidence these scores flow into targeting/audience activation.

   • [claimed-docs] “Profile Builder generates profiles based on the `traits` object in `identify` events.”
   • [claimed-docs] “You can also define custom logic for profile generation using a JavaScript function, allowing you to leverage up to a year’s worth of user e…”
   • [claimed-docs] “Profile Builder generates profiles based on the traits object in identify events. You can also define custom logic for profile generation us…”
   • [claimed-docs] “allowing you to leverage up to a year’s worth of user events data.”
   • [claimed-docs] “Profiles are customer records stored in your warehouse, based on the events data you send to Jitsu.”

   proof ↗⚑ flag

   Hightouch~partialclaimed5/10

   Hightouch clearly supports computed traits: models can be built with SQL, dbt, or a visual builder and used directly in audience/journey targeting via Customer Studio, and Identity Resolution can produce a golden-record profile view. However, there is no evidence of a native predictive-scoring capability (LTV, churn, purchase propensity) computed by Hightouch itself — the closest feature, 'AI Decisioning,' is a reinforcement-learning message/channel/timing optimizer, not a per-profile predictive score usable as an audience filter attribute. Missing for 10: documentation of built-in ML models producing LTV/churn/propensity scores stored as profile attributes, and evidence these specific score types are selectable in audience-builder filters.

   • [claimed-docs] “Models — Define reusable datasets with SQL, dbt, or a visual builder”
   • [claimed-docs] “Data teams typically create models by writing SQL in the editor, selecting from tables or views, or re-using dbt models and BI queries from …”
   • [claimed-docs] “Identity Resolution links related customer records into unified identities in your warehouse, so every downstream product works from one con…”
   • [claimed-docs] “Golden Record — a one-row-per-identity table of canonical field values (selected by survivorship rules) that many teams use as their source-…”
   • [claimed-docs] “AI Decisioning uses reinforcement learning to recommend the best message, channel, and timing for each customer, then measures what worked.”
   • [claimed-docs] “Create audiences from warehouse data, build automated journeys, and use the built-in agent to explore segments and analyze campaign performa…”
   • [claimed-docs] “Once the schema is configured, marketers can build audiences and journeys in a no-code UI.”

   proof ↗⚑ flag

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

   weight 2 · round to Hightouch
   Jitsu~partialclaimed4/10

   Jitsu supports batching of event delivery, deduplication of repeated sends, and connector syncs that pull large volumes of data into a warehouse, plus a CLI that manages multiple workspace config objects — all of which imply some bulk/batch data handling. However, there is no explicit bulk-operation API or CLI command for acting on many discrete items (e.g., bulk edit/delete of profiles, destinations, or events) beyond individual config management. Missing for 10: a documented bulk API/CLI command operating on many items at once, evidence of batch profile/record updates, and independent confirmation of bulk workflows in practice.

   • [claimed-docs] “Sub-second speed of data delivery to your warehouse. Optional batching is also supported”
   • [claimed-docs] “Send same data as many times as you want. Jitsu will deduplicate it for you”
   • [github] “Connector syncs — pull data *into* your warehouse from third-party sources (Airbyte-compatible connectors).”
   • [claimed-docs] “`jitsu-cli` is the command-line interface for Jitsu. It manages workspace configuration objects (destinations, streams, connections, …) and …”
   • [github] “jitsu-cli init scaffolds a TypeScript project with tests, and jitsu-cli deploy ships it to your workspace.”

   proof ↗⚑ flag

   Hightouch~partialclaimed6/10

   Hightouch's core sync engine operates on whole models/audiences (bulk rows) on each run, and the platform explicitly supports batch event ingestion ("Use the batch endpoint to submit multiple events... in a single request") plus a REST API for programmatically managing syncs, models, sources, and destinations. However, evidence doesn't show explicit bulk-management API operations (e.g., batch-update many syncs/audiences in one call) — it only shows single-resource CRUD and batch data ingestion. Missing for 10: explicit bulk/batch API operations across multiple resources at once (not just event ingestion), and independent verification of large-scale bulk sync performance.

   • [claimed-docs] “Use the batch endpoint to submit multiple events to the collector in a single request.”
   • [claimed-docs] “Hightouch exposes a REST API that lets users interact with resources like syncs, models, sources, and destinations.”
   • [claimed-docs] “A sync connects one model to one destination object, list, or table, and describes how the model's rows map into that destination.”
   • [claimed-docs] “A sync runs on the schedule or trigger you set, then reports the outcome of each run so you can confirm it worked.”
   • [claimed-docs] “Git Sync lets you manage workspace resources programmatically.”

   proof ↗⚑ flag

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

   weight 3 · round drawn
   Jitsu~partialcommunity6/10

   Jitsu Functions let users write custom JS/TypeScript logic that runs automatically on every event to filter, transform, or enrich it before delivery, and Profile Builder supports custom JS logic on identify events — both qualify as rule-based automatic actions triggered by events. However, this is scoped to data-pipeline transformations rather than general-purpose conditional workflows (e.g., alerts, multi-step triggers, external API calls beyond enrichment), and an early community review noted functions were seen as less mature/absent compared to Segment's equivalent feature. Missing for 10: evidence of broader conditional/multi-action rule chains beyond filter/transform/enrich, and independent hands-on confirmation of Functions' reliability in production.

   • [claimed-docs] “Functions are designed to process data in Jitsu before it is sent to the destination... Filter... Transform... Enrich”
   • [claimed-docs] “Modify, filter or augment events before it is stored in your warehouse with Jitsu Functions.”
   • [claimed-docs] “Functions are designed to process data in Jitsu before it is sent to the destination. These functions, written in JavaScript, provide variou…”
   • [claimed-docs] “Jitsu comes with a functions debugger/editor that allows to run function on a sample data”
   • [claimed-docs] “Profile Builder generates profiles based on the `traits` object in `identify` events.”
   • [claimed-docs] “You can also define custom logic for profile generation using a JavaScript function, allowing you to leverage up to a year’s worth of user e…”
   • [community] “Functions are an underrated and highly powerful feature of Segment... It doesn't seem present here and that would hold a customer such as my…”

   proof ↗⚑ flag

   Hightouch~partialclaimed6/10

   Hightouch supports event-driven syncs (trigger on data change/CDC), sync alerts on failure, and journeys that react to customer behavior, which cover rule-based automation of actions on events; however these are framed around marketing/data-sync workflows rather than general-purpose event-condition-action rule definition for arbitrary AI-native automations. missing for 10: a documented general rules/conditions engine (if-this-then-that style) independent of syncs/journeys, evidence of arbitrary event types triggering arbitrary actions beyond syncing/audience creation, and independent/hands-on confirmation of trigger-based automation working reliably.

   • [claimed-docs] “A sync runs on the schedule or trigger you set, then reports the outcome of each run so you can confirm it worked.”
   • [claimed-docs] “Sync modeled data to sales, support, analytics, and internal tools on a schedule or when data changes.”
   • [claimed-docs] “Get notified when a run fails [Configure sync alerts]”
   • [claimed-docs] “Instead of sending the full query result on every run, CDC compares the current results to the previous run and only sends new, changed, and…”
   • [claimed-docs] “Create audiences from warehouse data, build automated journeys, and use the built-in agent to explore segments and analyze campaign performa…”

   proof ↗⚑ flag

3. ai-native userSchedule recurring jobs or workflows

   weight 2 · round to Hightouch
   Jitsu~partialclaimed4/10

   Jitsu's self-hosting docs mention that connector syncs run as Kubernetes CronJobs, implying a recurring scheduling mechanism for data pulls, but there is no evidence of a user- or AI-agent-facing API/UI to define, view, or manage custom recurring jobs/workflows beyond this infrastructure detail. Missing for 10: explicit scheduling API/config surface, AI-agent-driven schedule creation via MCP, and any documentation of configurable intervals or workflow orchestration beyond connector syncs.

   • [claimed-docs] “A feature-complete Jitsu deployment requires a Kubernetes cluster: functions and profile builders run in dedicated function-server deploymen…”
   • [github] “Connector syncs — pull data *into* your warehouse from third-party sources (Airbyte-compatible connectors).”
   • [github] “Connector syncs — pull data into your warehouse from third-party sources (Airbyte-compatible connectors).”
   • [claimed-docs] “Jitsu runs an MCP server, so AI agents can manage your pipeline directly: create destinations, wire up streams, inspect Live Events, and edi…”

   proof ↗⚑ flag

   Hightouch✓fullprobed7/10

   Hightouch syncs run on a defined schedule or trigger and can sync on a cadence or when data changes, and journeys can be built as automated workflows; the MCP server lets an AI assistant manage syncs (which include their schedules) and journeys in natural language. Missing for 10: an explicit worked example of an AI assistant creating or modifying a sync schedule via MCP/natural language (only audience/sync creation and ad generation are shown), and independent confirmation of AI-driven scheduling in practice.

   • [claimed-docs] “A sync runs on the schedule or trigger you set, then reports the outcome of each run so you can confirm it worked.”
   • [claimed-docs] “Sync modeled data to sales, support, analytics, and internal tools on a schedule or when data changes.”
   • [claimed-docs] “Create audiences from warehouse data, build automated journeys, and use the built-in agent to explore segments and analyze campaign performa…”
   • [claimed-docs] “the Hightouch MCP server connects your AI assistant directly to your marketing stack so your team can build audiences, manage syncs, design …”
   • [claimed-docs] “"Create an audience of customers that added an item to their cart in the last day but haven't completed a transaction, then create a sync to…”
   • [probe] “PROBE runtime (recorded 2026-09-08): hightouch.com publishes an agent-oriented llms.txt on its main origin ('# Hightouch — Hightouch is an A…”

   proof ↗⚑ flag

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

   weight 1 · round to Hightouch
   Jitsu~partialclaimed4/10

   Jitsu's CLI workflow (`jitsu-cli init`/`deploy` from your own repo) implies git-based versioning of Functions, and the audit log with account-activity alerts supports review, while the dead-letter queue lets failed events be replayed. However, there is no documented built-in UI for viewing version history or rolling back a Function/pipeline to a prior state — versioning relies on the user's own repo, not a native Jitsu feature. Missing for 10: native version history/diff view for Functions, an explicit one-click rollback mechanism for pipeline configs, and audit-trail linkage between config changes and rollbacks.

   • [github] “build and deploy them from your own repo with the Jitsu CLI — `jitsu-cli init` scaffolds a TypeScript project with tests, and `jitsu-cli dep…”
   • [claimed-docs] “Audit log (SOC2-oriented) with account-activity alerts, plus named user API tokens with expiration.”
   • [claimed-docs] “Dead-letter queue with a reprocessing worker: failed events can be replayed instead of lost.”
   • [claimed-docs] “jitsu-cli is the command-line interface for Jitsu. It manages workspace configuration objects (destinations, streams, connections, …) and ru…”

   proof ↗⚑ flag

   Hightouch~partialcommunity5/10

   Hightouch's Git Sync lets teams manage workspace resources (models, syncs, audiences) programmatically via git, and the platform separately supports approvals/ownership boundaries for review workflows, with a community comparison confirming 'git version control' as a differentiator. However, there is no explicit documentation of a rollback mechanism, version history UI, or how git-based changes propagate back to rollback a live automation. Missing for 10: explicit rollback/restore documentation, versioned change history in the UI, and independent confirmation that Git Sync supports full revert workflows for automations.

   • [claimed-docs] “Git Sync lets you manage workspace resources programmatically.”
   • [claimed-docs] “Set up permissions, approvals, and ownership boundaries so each team has the right level of access.”
   • [community] “Hightouch vs Census comparison from the company: Hightouch offers more developer-focused features (live debugger, alerting, git version cont…”

   proof ↗⚑ flag

Destinations

1. data engineerRoute events to a large catalog of documented destination integrations with per-destination mapping and filtering

   weight 3 · round to Hightouch
   Jitsu~partialcommunity6/10

   Jitsu documents delivery to major warehouses (ClickHouse, BigQuery, Snowflake, Redshift, Postgres, S3/GCS) plus 'dozens of SaaS tools' and a public destinations catalog page, with per-event filtering/transform/enrichment via Functions and per-source-to-destination connection control (jitsu-docs-30, jitsu-gh-2, jitsu-docs-45, jitsu-docs-3/25/31). However, hands-on community feedback flags a narrower/more effortful integration catalog than competitors like Segment and calls out missing granular per-event transformation power at the time of review (jitsu-comm-6, jitsu-comm-7), and no evidence quantifies catalog size or shows per-destination field-mapping docs beyond generic Functions. Missing for 10: a large documented catalog comparable to Segment/Fivetran scale, explicit per-destination field-mapping configuration docs, and independent confirmation that filtering/mapping works smoothly at scale.

   • [github] “Delivery to destinations — ClickHouse, BigQuery, Snowflake, Redshift, Postgres, S3, GCS, and dozens of SaaS tools, streamed or micro-batched…”
   • [claimed-docs] “Send data to one of those destinations with Jitsu. Learn more about destination on docs website”
   • [claimed-docs] “Functions are designed to process data in Jitsu before it is sent to the destination... Filter... Transform... Enrich”
   • [claimed-docs] “Modify, filter or augment events before it is stored in your warehouse with Jitsu Functions.”
   • [claimed-docs] “Filter, transform, and enrich events with TypeScript functions before they land.”
   • [claimed-docs] “You can control which events become active by choosing not to connect a source to a destination or by using a Jitsu Function to filter out u…”
   • [community] “You've got most major ads sources that I care about, but it seems that there is a higher bar to implementation. Segment lets me just plug in…”
   • [community] “Functions are an underrated and highly powerful feature of Segment... It doesn't seem present here and that would hold a customer such as my…”

   proof ↗⚑ flag

   Hightouch✓fullcommunity7/10

   Hightouch's docs describe syncs that map one model to a destination object/list/table with per-destination field mapping, unlimited destination counts, scheduled/CDC-based delivery, and run-level debugging, and a HN co-founder interview independently confirms declarative field mapping plus a claim of 70+ deep integrations with visual audience filtering versus competitors. Missing for 10: a public, browsable catalog page listing/counting all supported destinations, and detailed docs on per-destination filtering syntax beyond audience-level filtering.

   • [claimed-docs] “A sync connects one model to one destination object, list, or table, and describes how the model's rows map into that destination.”
   • [claimed-docs] “Sync modeled data to sales, support, analytics, and internal tools on a schedule or when data changes.”
   • [claimed-docs] “Up to 2 active syncs * Unlimited destination count * Unlimited user seats”
   • [claimed-docs] “Instead of sending the full query result on every run, CDC compares the current results to the previous run and only sends new, changed, and…”
   • [claimed-docs] “Get notified when a run fails [Configure sync alerts]”
   • [community] “Hightouch cofounder explained syncs are declarative: you define a SQL model, and Hightouch figures out how to map it to the destination (e.g…”
   • [community] “Hightouch vs Census comparison from the company: Hightouch offers more developer-focused features (live debugger, alerting, git version cont…”

   proof ↗⚑ flag

Ingest

1. data engineerSend events server-to-server through a documented HTTP ingestion API with authentication and delivery guarantees

   weight 2 · round to Jitsu
   Jitsu✓fullprobed8/10

   Jitsu documents an HTTP ingestion API for server-to-server sending [jitsu-docs-2], with delivery guarantees such as deduplication, buffering to Kafka during warehouse downtime, and sub-second/batched delivery [jitsu-docs-7, jitsu-docs-8, jitsu-docs-9], plus a dead-letter queue with reprocessing for failed events [jitsu-docs-16]. Security/auth is covered via TLS/AES-256 encryption, SOC2 compliance, and named API tokens with expiration [jitsu-docs-22, jitsu-docs-23, jitsu-docs-17]. Missing for 10: a formal OpenAPI/API reference spec (probe found only 404s for openapi.json endpoints) and explicit documentation of per-request authentication mechanics for the HTTP ingestion endpoint itself.

   • [claimed-docs] “You can use HTTP API to send data to Jitsu. This is useful if you want to send data from your backend.”
   • [claimed-docs] “Sub-second speed of data delivery to your warehouse. Optional batching is also supported”
   • [claimed-docs] “Put your data warehouse in a maintenance mode? Jitsu will keep your data in a Kafka and deliver it once the warehouse is back online”
   • [claimed-docs] “Send same data as many times as you want. Jitsu will deduplicate it for you”
   • [claimed-docs] “Dead-letter queue with a reprocessing worker: failed events can be replayed instead of lost.”
   • [claimed-docs] “Audit log (SOC2-oriented) with account-activity alerts, plus named user API tokens with expiration.”
   • [claimed-docs] “All customer data is encrypted at rest with AES-256 and in transit via TLS.”
   • [claimed-docs] “Jitsu is SOC 2 compliant, demonstrating our commitment to maintaining the highest standards of security and data protection.”
   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.jitsu.com/openapi.json, https://docs.jitsu.com/swagger.json, https://docs.jitsu.com/api…”

   proof ↗⚑ flag

   Hightouch~partialclaimed6/10

   Hightouch documents an HTTP Tracking API for server-to-server event ingestion, including a batch endpoint and defined event types (identify, track, page, screen, group), which lands in the customer's warehouse. However, the evidence does not explicitly document authentication requirements or delivery guarantees (retries, idempotency, at-least-once semantics) specifically for this Events API — the bearer-token auth and run-level failure alerts/debugging described elsewhere apply to the general REST API and sync pipeline, not confirmed for the events collector itself. missing for 10: explicit auth mechanism for the HTTP Tracking/Events API, documented delivery/retry guarantees or acknowledgement semantics for event ingestion, and independent confirmation of reliability at scale.

   • [claimed-docs] “Hightouch provides SDKs for browser, iOS, Android, Node.js, and other platforms, plus an HTTP API for server-side ingestion”
   • [claimed-docs] “Hightouch's HTTP Tracking API enables you to record data from any website or application.”
   • [claimed-docs] “Hightouch provides SDKs for [browser](/docs/events/sdks/browser), [iOS](/docs/events/sdks/ios), [Android](/docs/events/sdks/android), [Node.…”
   • [claimed-docs] “Use the batch endpoint to submit multiple events to the collector in a single request.”
   • [claimed-docs] “Events come in five types — identify, track, page, screen, and group — each for a different kind of action.”
   • [claimed-docs] “You instrument your app, site, or backend with a Hightouch SDK, the HTTP API, or a streaming source like Kafka or Google Pub/Sub.”
   • [claimed-docs] “Authenticate each request by passing your API key as a bearer token in the `Authorization` header”

   proof ↗⚑ flag

Sdks

1. data engineerCollect events from web, mobile, and server apps through official SDKs that implement a documented tracking spec (track, identify, page)

   weight 3 · round to Hightouch
   Jitsu~partialclaimed6/10

   Jitsu documents a JS SDK that is '100% compatible with Segment API' (implying track/identify/page methods), a React integration, and an HTTP API for server-side event sends, plus identity-stitching and traits-based identify handling — covering web and server collection with a documented spec-like interface. However, there is no evidence of official mobile SDKs (iOS/Android) or a standalone published tracking-spec document beyond the Segment-compatibility claim. Missing for 10: dedicated mobile SDKs, explicit documented track/identify/page API reference (vs. inferred Segment compatibility), independent hands-on confirmation of SDK behavior across platforms.

   • [claimed-docs] “Jitsu javascript integration is 100% compatible with Segment API.”
   • [claimed-docs] “You can use HTTP API to send data to Jitsu. This is useful if you want to send data from your backend.”
   • [claimed-docs] “Once the user identity is known, Jitsu automatically amends all previous records associated with this user based on anonymous cookie-based i…”
   • [claimed-docs] “Profile Builder generates profiles based on the `traits` object in `identify` events.”
   • [claimed-docs] “Jitsu javascript integration is 100% compatible with Segment API. In addition to that we have React integration”
   • [claimed-docs] “select * from events where context_page_path = '/signup' and timestamp > now() - interval '1 day'”

   proof ↗⚑ flag

   Hightouch✓fullclaimed8/10

   Hightouch Events documents official SDKs for browser, iOS, Android, and Node.js plus an HTTP API/batch endpoint for server-side ingestion, and explicitly defines a tracking spec with five event types (identify, track, page, screen, group), matching the story's core ask. missing for 10: independent/hands-on developer corroboration that the SDKs work as documented in production, and no visible changelog/version history proving spec stability over time.

   • [claimed-docs] “Hightouch Events collects customer behavior from websites, mobile apps, and backend systems, then stores that data in your warehouse.”
   • [claimed-docs] “Hightouch provides SDKs for browser, iOS, Android, Node.js, and other platforms, plus an HTTP API for server-side ingestion”
   • [claimed-docs] “Hightouch's HTTP Tracking API enables you to record data from any website or application.”
   • [claimed-docs] “Events come in five types — identify, track, page, screen, and group — each for a different kind of action.”
   • [claimed-docs] “Hightouch provides SDKs for [browser](/docs/events/sdks/browser), [iOS](/docs/events/sdks/ios), [Android](/docs/events/sdks/android), [Node.…”
   • [claimed-docs] “Use the batch endpoint to submit multiple events to the collector in a single request.”
   • [claimed-docs] “You instrument your app, site, or backend with a Hightouch SDK, the HTTP API, or a streaming source like Kafka or Google Pub/Sub.”

   proof ↗⚑ flag

Sources

1. data engineerPull customer data in from third-party cloud apps and feeds — not just my own instrumented apps

   weight 1 · round to Jitsu
   Jitsu~partialcommunity5/10

   Jitsu explicitly advertises Airbyte-compatible 'Connector syncs' that pull data into the warehouse from third-party sources, directly matching the story, but this is only a single GitHub README line without documentation depth on which apps/feeds are supported. A community commenter also notes Jitsu has a 'higher bar to implementation' for pulling ad-platform data compared to Segment, suggesting real friction in practice. Missing for 10: detailed docs/connector catalog, first-party walkthrough of connector setup, and stronger independent corroboration that third-party pulls work smoothly.

   • [github] “Connector syncs — pull data *into* your warehouse from third-party sources (Airbyte-compatible connectors).”
   • [github] “Connector syncs — pull data into your warehouse from third-party sources (Airbyte-compatible connectors).”
   • [community] “You've got most major ads sources that I care about, but it seems that there is a higher bar to implementation. Segment lets me just plug in…”

   proof ↗⚑ flag

   Hightouch—none0/10

   Hightouch's evidence describes a reverse-ETL/composable-CDP architecture: models are read FROM the warehouse and synced OUT to destinations (docs-2, docs-20, docs-27), and its only inbound 'ingestion' feature (Hightouch Events) explicitly covers only the customer's own instrumented websites/apps/backends via SDKs or HTTP API (docs-5, docs-6, docs-41) — the exact case the story excludes. There is no documented source connector for pulling data in from third-party cloud apps/feeds (e.g., Salesforce, Stripe, Zendesk) into the warehouse; 'sources' in the API refer to warehouses, not external SaaS ingestion.

   • [claimed-docs] “Data teams typically create models by writing SQL in the editor, selecting from tables or views, or re-using dbt models and BI queries from …”
   • [claimed-docs] “Hightouch Events collects customer behavior from websites, mobile apps, and backend systems, then stores that data in your warehouse.”
   • [claimed-docs] “Hightouch provides SDKs for browser, iOS, Android, Node.js, and other platforms, plus an HTTP API for server-side ingestion”
   • [claimed-docs] “You instrument your app, site, or backend with a Hightouch SDK, the HTTP API, or a streaming source like Kafka or Google Pub/Sub.”
   • [claimed-docs] “A sync connects one model to one destination object, list, or table, and describes how the model's rows map into that destination.”
   • [claimed-docs] “Sync modeled data to sales, support, analytics, and internal tools on a schedule or when data changes.”

   proof ↗⚑ flag

Profiles

1. data engineerQuery unified customer profiles — traits, identifiers, event history — through a documented profile API or store

   weight 2 · round to Hightouch
   Jitsu~partialprobed5/10

   Jitsu documents a Profile Builder that generates unified customer profiles (traits, event history up to a year) stored in the warehouse, and profiles are queryable via SQL since they land in the customer's own warehouse tables, plus a Functions API (ctx.getWarehouse) to query warehouse data. However, there is no documented dedicated 'Profile API' or profile store/endpoint — access is only via the underlying warehouse SQL, and no OpenAPI/API reference for profiles was found (openapi probes 404). missing for 10: a documented profile-specific query API or SDK, dedicated identifier-resolution endpoints, independent/hands-on confirmation of profile querying beyond vendor docs.

   • [claimed-docs] “Profile Builder generates profiles based on the `traits` object in `identify` events.”
   • [claimed-docs] “You can also define custom logic for profile generation using a JavaScript function, allowing you to leverage up to a year’s worth of user e…”
   • [claimed-docs] “Profile Builder generates profiles based on the traits object in identify events. You can also define custom logic for profile generation us…”
   • [claimed-docs] “Profiles are customer records stored in your warehouse, based on the events data you send to Jitsu.”
   • [claimed-docs] “`ctx.getWarehouse` - warehouse API to query your data warehouses”
   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.jitsu.com/openapi.json, https://docs.jitsu.com/swagger.json, https://docs.jitsu.com/api…”

   proof ↗⚑ flag

   Hightouch~partialprobed6/10

   Hightouch's Identity Resolution builds a warehouse-resident Golden Record (traits, canonical identifiers) that docs explicitly say teams can 'query directly' or build models on (hightouch-docs-9, hightouch-docs-43), and its hybrid architecture keeps all profile data in the customer's own warehouse/store rather than a proprietary store (hightouch-comm-3). However, this satisfies the 'store' half of the story via SQL access to warehouse tables rather than a dedicated, documented Profile API for identifier/trait/event lookups — the REST API documented (hightouch-docs-13, hightouch-probe-rt-1) covers syncs/models/sources/destinations management, not profile/event retrieval by customer ID. Missing for 10: a first-party 'Profile API' endpoint (e.g., lookup-by-identifier returning traits+event history), explicit event-history query capability via API, and independent/hands-on confirmation of querying golden-record data through an API rather than raw SQL.

   • [claimed-docs] “Identity Resolution links related customer records into unified identities in your warehouse, so every downstream product works from one con…”
   • [claimed-docs] “Golden Record — a one-row-per-identity table of canonical field values (selected by survivorship rules) that many teams use as their source-…”
   • [claimed-docs] “Query these directly, build models on them in Customer Studio, or sync them downstream.”
   • [claimed-docs] “Hightouch exposes a REST API that lets users interact with resources like syncs, models, sources, and destinations.”
   • [community] “Hightouch's hybrid architecture stores no data in Hightouch itself out of the box—it all stays in the customer's own warehouse/S3 bucket—whi…”
   • [probe] “PROBE runtime (recorded 2026-09-08): Hightouch's REST API is live and cleanly auth-gated — a keyless GET https://api.hightouch.com/api/v1/sy…”

   proof ↗⚑ flag

Stitching

1. data engineerAnonymous and known activity stitches into one customer profile across devices, with documented and configurable identity-resolution rules

   weight 3 · round to Hightouch
   Jitsu~partialclaimed5/10

   Jitsu documents automatic identity stitching that amends prior anonymous records once a user is identified via cookie-based ID, and Profile Builder allows custom JS logic for profile generation, which gives some configurability. However there is no documented cross-device merging mechanism (only cookie-based, single-device) and no explicit documented rule set for identity resolution beyond the default cookie linkage. Missing for 10: cross-device identity linkage evidence, explicit documented/configurable rules for merging identities (e.g. via user ID mapping across devices), independent verification of stitching accuracy.

   • [claimed-docs] “Once the user identity is known, Jitsu automatically amends all previous records associated with this user based on anonymous cookie-based i…”
   • [claimed-docs] “Profile Builder generates profiles based on the `traits` object in `identify` events.”
   • [claimed-docs] “You can also define custom logic for profile generation using a JavaScript function, allowing you to leverage up to a year’s worth of user e…”
   • [claimed-docs] “Profile Builder generates profiles based on the traits object in identify events. You can also define custom logic for profile generation us…”
   • [claimed-docs] “Profiles are customer records stored in your warehouse, based on the events data you send to Jitsu.”

   proof ↗⚑ flag

   Hightouch~partialclaimed6/10

   Hightouch documents Identity Resolution linking cross-device/channel signals into a unified identity with a Golden Record built via survivorship rules, and describes unifying data across sources — directly addressing stitching known/anonymous activity into one profile. However, the docs shown don't detail configurable matching rules (e.g., deterministic vs probabilistic matching logic, field-level match keys) beyond survivorship for golden records, and there's no independent/hands-on validation of identity-resolution accuracy. Missing for 10: documentation of specific configurable match-rule types (deterministic/probabilistic, custom match keys), worked examples of anonymous-to-known stitching, and independent verification of resolution quality.

   • [claimed-docs] “Identity Resolution links related customer records into unified identities in your warehouse, so every downstream product works from one con…”
   • [claimed-docs] “Golden Record — a one-row-per-identity table of canonical field values (selected by survivorship rules) that many teams use as their source-…”
   • [claimed-docs] “Resolve identities across devices and channels so teams can build on a more complete customer record.”
   • [claimed-docs] “Unify customer or account data across multiple sources”
   • [claimed-docs] “Identity Resolution links these signals together so related records are treated as a single identity.”
   • [claimed-docs] “Query these directly, build models on them in Customer Studio, or sync them downstream.”
   • [claimed-docs] “If you're coming from a tool that stored and served customer profiles for you, this is the largest shift to plan for. The logic that lived i…”

   proof ↗⚑ flag

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

   weight 2 · round drawn
   Jitsu~partialprobed6/10

   Jitsu exposes substantial programmatic control beyond just data ingestion: an HTTP API for sending events, an official jitsu-cli for managing destinations/streams/connections and running the Functions dev workflow, and a first-party MCP server that lets AI agents create destinations, wire streams, inspect Live Events, and edit Functions — covering much of what the UI does. However, a direct probe for a public OpenAPI/REST management spec came back 404 on all candidate paths, suggesting no single comprehensive API surface documented for all UI actions (e.g., security/audit settings, billing, account admin). Missing for 10: a documented general-purpose REST/OpenAPI management API covering every UI screen (not just CLI/MCP-mediated actions), and independent confirmation that CLI/MCP truly reach full UI parity.

   • [claimed-docs] “You can use HTTP API to send data to Jitsu. This is useful if you want to send data from your backend.”
   • [claimed-docs] “`jitsu-cli` is the command-line interface for Jitsu. It manages workspace configuration objects (destinations, streams, connections, …) and …”
   • [github] “`jitsu-cli init` scaffolds a TypeScript project with tests, and `jitsu-cli deploy` ships it to your workspace.”
   • [claimed-docs] “Jitsu runs an MCP server, so AI agents can manage your pipeline directly: create destinations, wire up streams, inspect Live Events, and edi…”
   • [probe] “official MCP server documented at https://jitsu.com/docs/mcp”
   • [probe] “official CLI documented at https://jitsu.com/docs/jitsu-cli”
   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.jitsu.com/openapi.json, https://docs.jitsu.com/swagger.json, https://docs.jitsu.com/api…”
   • [probe] “PROBE runtime (recorded 2026-09-08): Jitsu's hosted MCP server ('Jitsu runs an MCP server, so AI agents can manage your pipeline directly: c…”

   proof ↗⚑ flag

   Hightouch~partialprobed6/10

   Hightouch documents a real REST API covering syncs, models, sources, and destinations (hightouch-docs-13, hightouch-probe-rt-1) and a separate MCP server that lets an AI assistant build audiences, journeys, syncs, and creatives in natural language (hightouch-docs-10/11/12, hightouch-probe-4). However, core UI-only features like audience/journey building are explicitly described as done in a 'no-code UI' (hightouch-docs-1/24/35), and the plain REST API guide never lists audience or journey endpoints, only syncs/models/sources/destinations — so parity between UI and raw API is not fully documented; the MCP path partially closes this gap but is a distinct conversational layer, not the generic API. Missing for 10: explicit REST API endpoints for audience/journey creation, and confirmation that MCP/API actions cover 100% of UI capabilities (e.g., Ad Studio, AI Decisioning) rather than a subset.

   • [claimed-docs] “marketers can build audiences and journeys in a no-code UI. Audiences cannot be created until this step is complete.”
   • [claimed-docs] “Hightouch exposes a REST API that lets users interact with resources like syncs, models, sources, and destinations.”
   • [claimed-docs] “the Hightouch MCP server connects your AI assistant directly to your marketing stack so your team can build audiences, manage syncs, design …”
   • [claimed-docs] “"Create an audience of customers that added an item to their cart in the last day but haven't completed a transaction, then create a sync to…”
   • [claimed-docs] “Once the schema is configured, marketers can build audiences and journeys in a no-code UI.”
   • [probe] “official MCP server documented at https://hightouch.com/docs/ai-integrations/mcp”
   • [probe] “PROBE runtime (recorded 2026-09-08): Hightouch's REST API is live and cleanly auth-gated — a keyless GET https://api.hightouch.com/api/v1/sy…”

   proof ↗⚑ flag

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

   weight 3 · round to Jitsu
   Jitsu~partialprobed6/10

   Jitsu is open-source and self-hostable (MIT license, Kubernetes/Helm/docker-compose deployments), and data lands directly in your own warehouse (ClickHouse, BigQuery, Snowflake, Postgres, S3) in standard formats, which supports data portability and exit. However, there is no explicit documented 'export all your data and leave' workflow, no bulk export/backup tool, and no discussion of exporting configuration/pipeline definitions or profiles data in an open interchange format beyond what lives in the warehouse. missing for 10: an explicit data-export/backup feature or docs describing full data portability/migration-out process, evidence on exporting Profile Builder or Function configs, independent confirmation of a clean full-data exit path.

   • [claimed-docs] “Run Jitsu yourself on Kubernetes with Helm, fully under your control.”
   • [claimed-docs] “Make your data warehouse the single source of truth for your data.”
   • [claimed-docs] “Make your data warehouse the single source of truth for your data. Jitsu is tailor-made for delivering the data to data warehouses as fast a…”
   • [github] “Delivery to destinations — ClickHouse, BigQuery, Snowflake, Redshift, Postgres, S3, GCS, and dozens of SaaS tools, streamed or micro-batched…”
   • [probe] “PROBE runtime (recorded 2026-09-08): Jitsu is really self-hostable from source — a keyless shallow clone of jitsucom/jitsu landed the MIT LI…”
   • [claimed-docs] “A feature-complete Jitsu deployment requires a Kubernetes cluster: functions and profile builders run in dedicated function-server deploymen…”

   proof ↗⚑ flag

   Hightouch~partialcommunity5/10

   Hightouch's hybrid architecture keeps customer data in the user's own warehouse/S3 by default rather than inside Hightouch (hightouch-comm-3), which means the underlying data is inherently in open, customer-controlled formats and not locked in. Git Sync and the REST API also let workspace resources (models, syncs) be managed/exported programmatically (hightouch-docs-32, hightouch-docs-13). Missing for 10: explicit first-party documentation of a full data export flow/format, no dedicated 'export and leave' guide, and this evidence comes mainly from a community HN thread rather than official docs.

   • [community] “Hightouch's hybrid architecture stores no data in Hightouch itself out of the box—it all stays in the customer's own warehouse/S3 bucket—whi…”
   • [claimed-docs] “Git Sync lets you manage workspace resources programmatically.”
   • [claimed-docs] “Hightouch exposes a REST API that lets users interact with resources like syncs, models, sources, and destinations.”

   proof ↗⚑ flag

3. ai-native userRead the product's source under an open license

   weight 2 · round to Jitsu
   Jitsu✓fullprobed8/10

   Jitsu's GitHub repo is confirmed to include the MIT LICENSE file, and probes verify the repo is publicly cloneable with self-host source (docker-compose.yml, source code) — plus docs and community threads describe it as an 'open-source data integration platform.' Missing for 10: no explicit license-file citation from claimed-docs pages and no independent legal/community discussion confirming license terms beyond the probe.

   • [probe] “PROBE runtime (recorded 2026-09-08): Jitsu is really self-hostable from source — a keyless shallow clone of jitsucom/jitsu landed the MIT LI…”
   • [probe] “PROBE llms.txt: HTTP 200 at https://docs.jitsu.com/llms.txt # Jitsu > Jitsu is an open-source data integration platform and a self-hostable…”
   • [community] “Congrats on the launch! We've started using the open source version for one of the tools we are building (CLI anonymized telemetry)... It wa…”

   proof ↗⚑ flag

   Hightouch—none0/10

   No evidence in the pack indicates Hightouch publishes its source code under any open license; all material describes a closed, hosted SaaS platform with commercial pricing tiers. The axis is fair for a data platform (open-core models exist in this space), but nothing here shows a public repo or license.

   ⚑ flag

4. ai-native userSelf-host the core product

   weight 3 · round to Jitsu
   Jitsu✓fullprobed8/10

   Jitsu is explicitly open source with documented self-hosting via Docker Compose and a full Kubernetes/Helm deployment path, confirmed by GitHub repo, docs, and an independent runtime probe showing the MIT-licensed repo and docker-compose.yml are actually accessible and cloneable. Community feedback corroborates real-world self-hosted deployment (HN commenter deployed OSS version to BigQuery), though one reviewer notes the Helm/deploy experience is rougher than competitors like Rudderstack. missing for 10: independent audit of feature-complete self-hosted parity (docs note some features need full K8s cluster), broader third-party validation of Helm chart quality beyond one critical comment.

   • [claimed-docs] “Run Jitsu yourself on Kubernetes with Helm, fully under your control.”
   • [claimed-docs] “Development Helm chart: deploy the full architecture to Minikube with zero configuration.”
   • [claimed-docs] “A feature-complete Jitsu deployment requires a Kubernetes cluster: functions and profile builders run in dedicated function-server deploymen…”
   • [probe] “PROBE runtime (recorded 2026-09-08): Jitsu is really self-hostable from source — a keyless shallow clone of jitsucom/jitsu landed the MIT LI…”
   • [community] “Congrats on the launch! We've started using the open source version for one of the tools we are building (CLI anonymized telemetry)... It wa…”
   • [community] “Rudderstack user here (and ex Segment)... rudderstack has deploy ready helm charts, which I'd argue are significantly better than docker com…”
   • [probe] “PROBE llms.txt: HTTP 200 at https://docs.jitsu.com/llms.txt # Jitsu > Jitsu is an open-source data integration platform and a self-hostable…”

   proof ↗⚑ flag

   Hightouch—none0/10

   Hightouch is presented throughout as a hosted SaaS platform (workspace, pricing tiers, API auth) with no mention of self-hosting or on-premise deployment options anywhere in the evidence pack.

   ⚑ flag

Observability

1. data engineerWatch events flow live and diagnose delivery failures per destination — debugger views, delivery metrics, and alerting

   weight 2 · round to Jitsu
   Jitsu~partialclaimed7/10

   Jitsu's Live Events view provides a debugger showing function logs, warehouse batch statuses, streaming delivery results, and dead-lettered events, with a dead-letter queue and reprocessing worker for failed deliveries, and this data can now be exported to an external monitoring stack near real-time (Datadog, Grafana Cloud, New Relic, Elastic via OTLP). However, native alerting is limited to SOC2-oriented account-activity audit alerts, not specific per-destination delivery-failure alert rules/thresholds built into Jitsu itself. Missing for 10: first-party alerting/paging on delivery failures within Jitsu, independent/hands-on confirmation of the debugger UI in practice, and per-destination metrics dashboards beyond changelog claims.

   • [claimed-docs] “Everything you can see in Live Events — function logs, warehouse batch statuses, streaming delivery results, and dead-lettered events — can …”
   • [claimed-docs] “Dead-letter queue with a reprocessing worker: failed events can be replayed instead of lost.”
   • [claimed-docs] “Records are delivered as OpenTelemetry (OTLP) log records over HTTP, so they work with Datadog, Grafana Cloud, New Relic, Elastic, or any Op…”
   • [claimed-docs] “Audit log (SOC2-oriented) with account-activity alerts, plus named user API tokens with expiration.”
   • [claimed-docs] “Jitsu comes with a functions debugger/editor that allows to run function on a sample data”

   proof ↗⚑ flag

   Hightouch~partialcommunity6/10

   Hightouch's docs confirm a sync-run debugger for checking runs and debugging rejected rows, plus configurable alerts when a run fails, and community evidence (comm-5) explicitly calls out 'live debugger, alerting' as differentiating developer-facing features versus competitors. However, there is no direct evidence of a live event-stream view (watching events flow in real time) or per-destination delivery metrics dashboards beyond the sync debugger and alerting. Missing for 10: live event flow visualization, per-destination delivery metrics/dashboards, and independent hands-on confirmation of debugger usability at scale.

   • [claimed-docs] “A sync runs on the schedule or trigger you set, then reports the outcome of each run so you can confirm it worked.”
   • [claimed-docs] “Check a run or debug rejected rows | Debug a sync run”
   • [claimed-docs] “Get notified when a run fails [Configure sync alerts]”
   • [claimed-docs] “Check a run or debug rejected rows | [Debug a sync run](/docs/syncs/debugger)”
   • [community] “Hightouch vs Census comparison from the company: Hightouch offers more developer-focused features (live debugger, alerting, git version cont…”

   proof ↗⚑ flag

Consent

1. privacy leadUser consent is captured and enforced across destinations — opt-outs and consent categories are honored downstream automatically

   weight 3 · round drawn
   Jitsu—none0/10

   No evidence anywhere in the pack describes consent capture, consent-category mapping, opt-out enforcement, or CMP/consent-signal propagation to destinations. Jitsu's Functions can filter events generically, but nothing documents this being used for consent enforcement, and no privacy/consent-specific feature is mentioned—only unrelated security/compliance items (SOC2, DPA, encryption).

   ⚑ flag

   Hightouch—none0/10

   No evidence pack items mention consent management, opt-outs, consent categories, or any suppression/enforcement mechanism honored across destinations; Hightouch's docs focus on syncs, audiences, identity resolution, and events, none of which describe consent capture or enforcement.

   ⚑ flag

Deletion

1. privacy leadProcess user deletion and suppression requests (GDPR/CCPA) and have them forwarded to connected destinations

   weight 2 · round drawn
   Jitsu—none0/10

   No evidence of a documented feature for processing GDPR/CCPA user deletion or suppression requests and propagating them to connected destinations; only general security/DPA compliance mentions exist, not a deletion workflow. Missing for 10: any documented delete/suppress user API, per-user erasure workflow, or downstream forwarding of deletion requests to destinations.

   • [claimed-docs] “All customer data is encrypted at rest with AES-256 and in transit via TLS.”
   • [claimed-docs] “Jitsu is SOC 2 compliant, demonstrating our commitment to maintaining the highest standards of security and data protection.”
   • [claimed-docs] “You can access DPA (Data Processing Agreement) + SCC (Standard Contractual Clause) using this link”

   proof ↗⚑ flag

   Hightouch—none0/10

   The evidence pack describes Hightouch's audience-building, sync, identity resolution, and events capabilities, but contains no mention of deletion/suppression request workflows, DSAR/GDPR-CCPA processing, or forwarding erasure requests to destinations. This is a plausible capability for a CDP/reverse-ETL platform, so the axis applies, but no evidence supports it.

   ⚑ flag

Pii controls

1. privacy leadControl PII flow per destination — hashing, masking, and field-level filtering of sensitive attributes

   weight 2 · round to Jitsu
   Jitsu~partialclaimed4/10

   Jitsu's Functions feature lets you write TypeScript/JavaScript to filter, transform, and enrich events before they reach a destination, and you can choose not to connect a source to a destination — this could be used to implement custom hashing/masking/field-filtering logic, giving privacy leads a mechanism for PII control. However, there's no dedicated, documented PII-specific feature (built-in hashing/masking primitives, a privacy policy UI, or per-destination redaction rules) — it's a general-purpose transform layer requiring custom code rather than a purpose-built privacy/consent control. missing for 10: built-in hashing/masking functions marketed for PII, per-destination privacy policy configuration, explicit field-level anonymization documentation or examples.

   • [claimed-docs] “Functions are designed to process data in Jitsu before it is sent to the destination... Filter... Transform... Enrich”
   • [claimed-docs] “Modify, filter or augment events before it is stored in your warehouse with Jitsu Functions.”
   • [claimed-docs] “Filter, transform, and enrich events with TypeScript functions before they land.”
   • [claimed-docs] “You can control which events become active by choosing not to connect a source to a destination or by using a Jitsu Function to filter out u…”
   • [claimed-docs] “Functions are designed to process data in Jitsu before it is sent to the destination. These functions, written in JavaScript, provide variou…”

   proof ↗⚑ flag

   Hightouch—none0/10

   The evidence pack covers Hightouch's sync engine, identity resolution, events, and permissions/RBAC (docs-34), but nowhere documents field-level hashing, masking, or PII-specific filtering controls scoped per destination. Since Hightouch moves warehouse data (including PII) to many destinations, this is a fair and plausible axis for a reverse-ETL/CDP product, but no capability is evidenced.

   ⚑ flag

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

   weight 2 · round drawn
   Jitsu~partialprobed5/10

   Jitsu can be fully self-hosted on your own Kubernetes/infrastructure (jitsu-docs-14, jitsu-docs-41, confirmed self-hostable in jitsu-probe-rt-2), which lets a user choose exactly where data is stored, and it offers DPA/SCC for compliance (jitsu-docs-44). However there's no documented region-selection feature for the hosted/cloud offering itself (e.g., 'choose EU vs US region' toggle) — self-hosting is the only mechanism for residency control. Missing for 10: explicit hosted-cloud region/residency picker, documentation of where hosted Jitsu Cloud data lives by default, and independent confirmation of residency guarantees beyond self-hosting.

   • [claimed-docs] “Run Jitsu yourself on Kubernetes with Helm, fully under your control.”
   • [claimed-docs] “A feature-complete Jitsu deployment requires a Kubernetes cluster: functions and profile builders run in dedicated function-server deploymen…”
   • [claimed-docs] “You can access DPA (Data Processing Agreement) + SCC (Standard Contractual Clause) using this link”
   • [probe] “PROBE runtime (recorded 2026-09-08): Jitsu is really self-hostable from source — a keyless shallow clone of jitsucom/jitsu landed the MIT LI…”

   proof ↗⚑ flag

   Hightouch~partialcommunity5/10

   Hightouch's hybrid architecture stores no customer data in Hightouch itself—data stays in the customer's own warehouse/S3 bucket—which gives users effective control over where their data resides by choosing their own warehouse's region, and this underpins its use by regulated fintech/healthcare customers. However, there is no explicit documentation of a region-selection setting or residency options within Hightouch's own platform/control plane. Missing for 10: explicit region/residency configuration UI or docs, statements about where Hightouch's own metadata/control plane is hosted, and any multi-region deployment options.

   • [community] “Hightouch's hybrid architecture stores no data in Hightouch itself out of the box—it all stays in the customer's own warehouse/S3 bucket—whi…”

   proof ↗⚑ flag

2. ai-native userControl data retention and deletion

   weight 2 · round to Jitsu
   Jitsu~partialclaimed4/10

   Jitsu offers self-hosting (full data ownership) and mentions DPA/SCC + SOC2 compliance, which implies some control over where/how data is stored, but there is no explicit retention-policy setting or data-deletion API/feature documented anywhere in the pack. Missing for 10: documented retention window controls, a delete/erase-user-data API or UI action, and any GDPR-style right-to-be-forgotten workflow.

   • [claimed-docs] “Run Jitsu yourself on Kubernetes with Helm, fully under your control.”
   • [claimed-docs] “A feature-complete Jitsu deployment requires a Kubernetes cluster: functions and profile builders run in dedicated function-server deploymen…”
   • [claimed-docs] “You can access DPA (Data Processing Agreement) + SCC (Standard Contractual Clause) using this link”
   • [claimed-docs] “Jitsu is SOC 2 compliant, demonstrating our commitment to maintaining the highest standards of security and data protection.”
   • [claimed-docs] “All customer data is encrypted at rest with AES-256 and in transit via TLS.”

   proof ↗⚑ flag

   Hightouch—none0/10

   The evidence pack has no documentation of data retention policies, deletion APIs, or AI-specific data lifecycle controls; the closest tangential fact is that Hightouch doesn't store customer data itself (hightouch-comm-3), but this doesn't describe any deletion/retention control mechanism for AI-native users.

   • [community] “Hightouch's hybrid architecture stores no data in Hightouch itself out of the box—it all stays in the customer's own warehouse/S3 bucket—whi…”

   proof ↗⚑ flag

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

   weight 2 · round drawn
   Jitsu—none0/10

   The evidence pack covers Jitsu's self-hosting, security/compliance, and data-pipeline features extensively, but contains no mention of an opt-out flag, setting, or documentation for disabling Jitsu's own product/CLI usage telemetry sent back to Jitsu Inc. Self-hosting (jitsu-docs-14, jitsu-docs-41) addresses customer analytics data sovereignty, not the product's own telemetry practices.

   • [claimed-docs] “Run Jitsu yourself on Kubernetes with Helm, fully under your control.”
   • [claimed-docs] “A feature-complete Jitsu deployment requires a Kubernetes cluster: functions and profile builders run in dedicated function-server deploymen…”
   • [claimed-docs] “All customer data is encrypted at rest with AES-256 and in transit via TLS.”
   • [claimed-docs] “Jitsu is SOC 2 compliant, demonstrating our commitment to maintaining the highest standards of security and data protection.”

   proof ↗⚑ flag

   Hightouch—none0/10

   No evidence pack content addresses telemetry/usage-tracking opt-out settings for Hightouch itself; all docs concern data sync, identity resolution, and API features unrelated to product telemetry controls.

   ⚑ flag

Replay

1. data engineerReplay archived events into a new destination or backfill history when a tool is added or a pipeline breaks

   weight 2 · round to Jitsu
   Jitsu~partialclaimed3/10

   Jitsu's dead-letter queue with a reprocessing worker lets failed events be replayed rather than lost, covering the 'pipeline breaks' half of the story, but there's no evidence of a mechanism to replay/backfill previously archived events into a newly added destination. missing for 10: documented backfill/replay tooling for historical events into new destinations, evidence of re-streaming warehouse-stored events, independent confirmation of DLQ reprocessing in practice.

   • [claimed-docs] “Dead-letter queue with a reprocessing worker: failed events can be replayed instead of lost.”
   • [claimed-docs] “Everything you can see in Live Events — function logs, warehouse batch statuses, streaming delivery results, and dead-lettered events — can …”

   proof ↗⚑ flag

   Hightouch—none0/10

   Hightouch's sync docs describe scheduled/triggered syncs, CDC, and run debugging, but there is no evidence of replaying archived events or backfilling historical data into a new destination when a pipeline breaks or a tool is added; Hightouch Events pipes live behavioral events into the warehouse, not archived-event replay. missing for 10: any mention of event/data replay, backfill mechanisms, historical resync into new destinations, or reprocessing archived data after pipeline failure.

   • [claimed-docs] “A sync runs on the schedule or trigger you set, then reports the outcome of each run so you can confirm it worked.”
   • [claimed-docs] “Instead of sending the full query result on every run, CDC compares the current results to the previous run and only sends new, changed, and…”
   • [claimed-docs] “For higher performance at scale, you can enable the Lightning sync engine to compute CDC directly in the warehouse.”
   • [claimed-docs] “Hightouch Events collects customer behavior from websites, mobile apps, and backend systems, then stores that data in your warehouse.”

   proof ↗⚑ flag

Data quality

1. data engineerA tracking plan or schema is enforced — violating events get flagged, blocked, or quarantined instead of silently corrupting downstream data

   weight 2 · round to Jitsu
   Jitsu~partialclaimed4/10

   Jitsu's Functions can filter, block, or transform events before they reach a destination, and a dead-letter queue captures failed events for reprocessing instead of silent loss — these are the building blocks a data engineer could use to implement custom validation. However, there is no documented tracking-plan/schema-enforcement feature (e.g., defining an event schema and auto-flagging/quarantining violations) — it would require building custom Function logic. Missing for 10: native schema/tracking-plan definition, automatic validation against that schema, and dedicated quarantine flagging distinct from generic DLQ failure handling.

   • [claimed-docs] “Functions are designed to process data in Jitsu before it is sent to the destination... Filter... Transform... Enrich”
   • [claimed-docs] “Modify, filter or augment events before it is stored in your warehouse with Jitsu Functions.”
   • [claimed-docs] “You can control which events become active by choosing not to connect a source to a destination or by using a Jitsu Function to filter out u…”
   • [claimed-docs] “Dead-letter queue with a reprocessing worker: failed events can be replayed instead of lost.”
   • [claimed-docs] “Functions are designed to process data in Jitsu before it is sent to the destination. These functions, written in JavaScript, provide variou…”

   proof ↗⚑ flag

   Hightouch—none0/10

   Hightouch's Events feature mentions a 'schema' step and syncs that can flag 'rejected rows' at the destination side, but there is no evidence of tracking-plan validation, blocking, or quarantining of malformed/violating events at ingestion — the debugger references relate to sync/destination failures, not schema enforcement on incoming event data.

   • [claimed-docs] “Once the schema is configured, marketers can build audiences and journeys in a no-code UI.”
   • [claimed-docs] “Once the schema is configured, marketers can build [audiences](/docs/customer-studio/usage) and [journeys](/docs/customer-studio/journeys) i…”
   • [claimed-docs] “Check a run or debug rejected rows | Debug a sync run”
   • [claimed-docs] “Check a run or debug rejected rows | [Debug a sync run](/docs/syncs/debugger)”
   • [claimed-docs] “Events come in five types — identify, track, page, screen, and group — each for a different kind of action.”

   proof ↗⚑ flag

Transformations

1. data engineerTransform, filter, and enrich events in-pipeline with custom code or functions before they reach destinations

   weight 2 · round to Jitsu
   Jitsu✓fullclaimed8/10

   Jitsu Functions are extensively documented as an in-pipeline transformation layer: filter/transform/enrich events in JavaScript/TypeScript before delivery, with a dev CLI workflow (jitsu-cli init/deploy), a functions debugger/editor to test on sample data, and pricing docs confirming you can filter events via Functions rather than routing rules. Missing for 10: independent hands-on validation of Functions in production (the only community feedback found is an old HN comment claiming Functions were absent at launch, which predates current docs and isn't a concrete recent contradiction) and no case-study evidence of complex enrichment logic in the wild.

   • [claimed-docs] “Functions are designed to process data in Jitsu before it is sent to the destination... Filter... Transform... Enrich”
   • [claimed-docs] “Modify, filter or augment events before it is stored in your warehouse with Jitsu Functions.”
   • [claimed-docs] “Filter, transform, and enrich events with TypeScript functions before they land.”
   • [claimed-docs] “Functions are designed to process data in Jitsu before it is sent to the destination. These functions, written in JavaScript, provide variou…”
   • [claimed-docs] “Jitsu comes with a functions debugger/editor that allows to run function on a sample data”
   • [claimed-docs] “You can control which events become active by choosing not to connect a source to a destination or by using a Jitsu Function to filter out u…”
   • [github] “`jitsu-cli init` scaffolds a TypeScript project with tests, and `jitsu-cli deploy` ships it to your workspace.”
   • [github] “jitsu-cli init scaffolds a TypeScript project with tests, and jitsu-cli deploy ships it to your workspace.”
   • [github] “build and deploy them from your own repo with the Jitsu CLI — `jitsu-cli init` scaffolds a TypeScript project with tests, and `jitsu-cli dep…”
   • [claimed-docs] “`ctx.getWarehouse` - warehouse API to query your data warehouses”

   proof ↗⚑ flag

   Hightouch~partialclaimed5/10

   Hightouch lets data engineers define models via SQL, dbt, or a visual builder (docs-2, docs-23), which supports filtering and transforming data before syncing to destinations, and CDC only sends changed rows (docs-25/36). However, there's no evidence of custom code/functions (e.g., JS/Python transform steps) applied in-pipeline for enrichment beyond SQL modeling, nor any dedicated transformation-step API distinct from model definition. Missing for 10: explicit custom-code/function transformation step, enrichment logic beyond SQL/dbt models, and independent verification of such capability.

   • [claimed-docs] “Data teams typically create models by writing SQL in the editor, selecting from tables or views, or re-using dbt models and BI queries from …”
   • [claimed-docs] “Models — Define reusable datasets with SQL, dbt, or a visual builder”
   • [claimed-docs] “Instead of sending the full query result on every run, CDC compares the current results to the previous run and only sends new, changed, and…”
   • [claimed-docs] “CDC compares the current results to the previous run and only sends new, changed, and removed rows.”

   proof ↗⚑ flag

Composable

1. data engineerRun warehouse-native: define models and audiences on tables already in my warehouse and activate them without re-collecting the data (reverse ETL)

   weight 2 · round to Hightouch
   Jitsu—none0/10

   Jitsu's Profile Builder explicitly builds profiles from events data sent *to* Jitsu ('Profiles are customer records stored in your warehouse, based on the events data you send to Jitsu'), not from pre-existing warehouse tables — the opposite of a warehouse-native, no-re-collection model. The only warehouse-read capability shown is `ctx.getWarehouse` inside Functions, which is far short of defining models/audiences on existing tables and activating them via reverse ETL. Missing for 10: any evidence of modeling on pre-existing warehouse tables, an audience-builder UI over arbitrary warehouse schemas, or activation/sync-back without first ingesting data through Jitsu's own collection pipeline.

   • [claimed-docs] “Profile Builder generates profiles based on the `traits` object in `identify` events.”
   • [claimed-docs] “Profiles are customer records stored in your warehouse, based on the events data you send to Jitsu.”
   • [claimed-docs] “`ctx.getWarehouse` - warehouse API to query your data warehouses”

   proof ↗⚑ flag

   Hightouch✓fullcommunity8/10

   Hightouch's core model is warehouse-native: models are defined via SQL/dbt/BI queries directly on warehouse tables (docs-2, docs-23), audiences/Customer Studio build on top of these warehouse models (docs-19, docs-43), and syncs activate data to destinations on a schedule without re-collecting or storing it elsewhere (docs-20, docs-27, docs-26 CDC computed in-warehouse). Independent community testimony corroborates that syncs are declarative SQL-to-destination mappings and that Hightouch stores no data itself, keeping everything in the customer's own warehouse (hightouch-comm-2, hightouch-comm-3). Missing for 10: independent hands-on performance benchmarks proving no-recollection at scale, and no third-party audit of the 'no data stored' claim beyond company/community anecdote.

   • [claimed-docs] “Data teams typically create models by writing SQL in the editor, selecting from tables or views, or re-using dbt models and BI queries from …”
   • [claimed-docs] “Models — Define reusable datasets with SQL, dbt, or a visual builder”
   • [claimed-docs] “Create audiences from warehouse data, build automated journeys, and use the built-in agent to explore segments and analyze campaign performa…”
   • [claimed-docs] “Sync modeled data to sales, support, analytics, and internal tools on a schedule or when data changes.”
   • [claimed-docs] “A sync connects one model to one destination object, list, or table, and describes how the model's rows map into that destination.”
   • [claimed-docs] “For higher performance at scale, you can enable the Lightning sync engine to compute CDC directly in the warehouse.”
   • [claimed-docs] “Query these directly, build models on them in Customer Studio, or sync them downstream.”
   • [community] “Hightouch cofounder explained syncs are declarative: you define a SQL model, and Hightouch figures out how to map it to the destination (e.g…”
   • [community] “Hightouch's hybrid architecture stores no data in Hightouch itself out of the box—it all stays in the customer's own warehouse/S3 bucket—whi…”

   proof ↗⚑ flag

Warehouse sync

1. data engineerRaw events and profiles land in my own warehouse or lake (Snowflake, BigQuery, ClickHouse, S3) on a schedule I control

   weight 3 · round to Jitsu
   Jitsu✓fullclaimed9/10

   Jitsu explicitly delivers events (and Profiles) to warehouses/lakes including Snowflake, BigQuery, ClickHouse, Redshift, Postgres, S3, GCS, with both real-time streaming and optional batching, dedup, auto-schema creation, and self-hosting options that give engineers full control over the delivery cadence and infrastructure. GitHub and docs corroborate the destination list and delivery modes, and self-hosting via Helm/Kubernetes reinforces 'own warehouse' control. Missing for 10: independent/hands-on confirmation of exact scheduling granularity (batch interval configuration specifics) and clearer documentation on how 'schedule I control' maps to concrete batch/cron settings.

   • [claimed-docs] “Sub-second speed of data delivery to your warehouse. Optional batching is also supported”
   • [claimed-docs] “Put your data warehouse in a maintenance mode? Jitsu will keep your data in a Kafka and deliver it once the warehouse is back online”
   • [claimed-docs] “Send same data as many times as you want. Jitsu will deduplicate it for you”
   • [claimed-docs] “Jitsu will automatically create tables and columns in your warehouse based on the data you send”
   • [github] “Delivery to destinations — ClickHouse, BigQuery, Snowflake, Redshift, Postgres, S3, GCS, and dozens of SaaS tools, streamed or micro-batched…”
   • [claimed-docs] “Make your data warehouse the single source of truth for your data.”
   • [claimed-docs] “Make your data warehouse the single source of truth for your data. Jitsu is tailor-made for delivering the data to data warehouses as fast a…”
   • [claimed-docs] “Run Jitsu yourself on Kubernetes with Helm, fully under your control.”
   • [claimed-docs] “A feature-complete Jitsu deployment requires a Kubernetes cluster: functions and profile builders run in dedicated function-server deploymen…”
   • [claimed-docs] “Profiles are customer records stored in your warehouse, based on the events data you send to Jitsu.”

   proof ↗⚑ flag

   Hightouch✓fullcommunity7/10

   Hightouch Events explicitly ingests raw behavioral events via SDKs, HTTP API, or streaming sources (Kafka/Pub/Sub) and stores them directly in the customer's own warehouse (Snowflake, BigQuery, etc.), and Identity Resolution builds unified 'golden record' profiles in that same warehouse — confirmed independently by an HN comment noting Hightouch's hybrid architecture stores no data itself, everything stays in the customer's warehouse/S3. Models are built with SQL/dbt directly against warehouse tables, keeping engineers in control of the data layer. missing for 10: explicit documentation of a configurable ingestion schedule/cadence for raw event landing (docs mostly describe real-time streaming ingestion, not a data-engineer-set batch schedule), and independent hands-on validation of warehouse-landing reliability at scale.

   • [claimed-docs] “Hightouch Events collects customer behavior from websites, mobile apps, and backend systems, then stores that data in your warehouse.”
   • [claimed-docs] “Hightouch provides SDKs for browser, iOS, Android, Node.js, and other platforms, plus an HTTP API for server-side ingestion”
   • [claimed-docs] “Hightouch's HTTP Tracking API enables you to record data from any website or application.”
   • [claimed-docs] “Hightouch provides SDKs for [browser](/docs/events/sdks/browser), [iOS](/docs/events/sdks/ios), [Android](/docs/events/sdks/android), [Node.…”
   • [claimed-docs] “You instrument your app, site, or backend with a Hightouch SDK, the HTTP API, or a streaming source like Kafka or Google Pub/Sub.”
   • [claimed-docs] “Identity Resolution links related customer records into unified identities in your warehouse, so every downstream product works from one con…”
   • [claimed-docs] “Golden Record — a one-row-per-identity table of canonical field values (selected by survivorship rules) that many teams use as their source-…”
   • [claimed-docs] “Models — Define reusable datasets with SQL, dbt, or a visual builder”
   • [community] “Hightouch's hybrid architecture stores no data in Hightouch itself out of the box—it all stays in the customer's own warehouse/S3 bucket—whi…”

   proof ↗⚑ flag