Bolt.new vs Replit

Agent access

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

   weight 2 · round drawn
   Bolt.new✓fullprobed8/10

   A direct probe confirms Bolt's support site serves a valid llms.txt file (HTTP 200) listing structured links to its docs, and the docs themselves are exposed in clean .md format suitable for agent consumption. This is concrete first-party evidence the capability works as claimed, not just a vendor claim. Missing for 10: no independent/community report of an agent actually consuming llms.txt in practice, and no evidence of additional agent-specific manifest formats beyond llms.txt.

   • [probe] “PROBE llms.txt: HTTP 200 at https://support.bolt.new/llms.txt # Bolt - [Intro to Bolt](https://support.bolt.new/get-started/intro-bolt.md):…”
   • [claimed-docs] “edit your code directly in code view”

   proof ↗⚑ flag

   Replit✓fullprobed8/10

   A probe confirms Replit docs serve a working llms.txt at docs.replit.com/llms.txt (HTTP 200) listing structured doc sections, directly enabling an agent to be pointed at agent-oriented docs; this is corroborated by extensive first-party docs pages formatted for agent consumption. Missing for 10: independent third-party confirmation that agents actually consume/use this llms.txt effectively, and no evidence of an OpenAPI spec or other machine-readable API docs (all probed paths 404).

   • [probe] “PROBE llms.txt: HTTP 200 at https://docs.replit.com/llms.txt # Replit - [Features (218 pages)](https://docs.replit.com/_llms/en/features.md…”
   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.replit.com/openapi.json, https://docs.replit.com/swagger.json, https://docs.replit.com/…”
   • [claimed-docs] “Replit Agent can connect to hundreds of external tools and data sources through the Model Context Protocol (MCP)”

   proof ↗⚑ flag

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

   weight 2 · round drawn
   Bolt.new—none0/10

   Bolt.new is a browser-based, chat/UI-driven builder; no evidence of a CLI, API, or headless mode for CI/automation—OpenAPI probes 404 and docs describe only interactive web workflows (chat, code view, collaborate, publish). No mention of scriptable or non-interactive invocation exists in the evidence pack.

   • [probe] “PROBE openapi: all candidate paths 404 (https://support.bolt.new/openapi.json, https://support.bolt.new/swagger.json, https://support.bolt.n…”
   • [claimed-docs] “edit your code directly in code view”
   • [claimed-docs] “Bolt offers two agents, Standard and Max, powered by large language models.”

   proof ↗⚑ flag

   Replit—none0/10

   Evidence covers Replit's Agent modes, MCP connectors, and scheduled 'routines' but nothing about a CLI, API, or headless mode for running Replit in CI/automation pipelines outside its own UI. missing for 10: CLI/API for headless execution, CI/CD pipeline integration, documented automation outside Replit's own scheduler.

   • [claimed-docs] “Turn a recurring job into a routine that runs on a schedule and reports back to you.”
   • [claimed-docs] “Agent Modes let you choose the balance between included allowance, credits, and capability before you send a request.”
   • [claimed-docs] “Choose Free Mode, Power Mode, or Max Mode based on the work you want to do, the capability you need, and your available allowance or credits…”

   proof ↗⚑ flag

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

   weight 3 · round to Replit
   Bolt.new~partialclaimed5/10

   Bolt's docs mention a 'connect-mcp' page and a built-in GitHub MCP connector for searching code, issues, and discussions (bolt-docs-18, bolt-docs-40), showing MCP tool-use exists, but there's no evidence users can add arbitrary/custom MCP servers beyond the GitHub connector. missing for 10: documentation of adding custom/third-party MCP servers, evidence of general MCP server registration UI, and independent confirmation of multi-server support.

   • [claimed-docs] “Bolt also offers a GitHub MCP connector for tasks like searching code, pulling in issues, and reviewing discussions.”
   • [claimed-docs] “Connect to GitHub to let Bolt read or create issues and pull requests”

   proof ↗⚑ flag

   Replit✓fullclaimed8/10

   Replit's docs explicitly describe connecting Agent to external tools via MCP, with a curated catalog (Stripe, Linear, Notion, Sentry) plus the ability to add custom servers by URL, giving Agent access to their tools. missing for 10: independent/hands-on community corroboration of MCP usage in practice, and detail on tool invocation reliability beyond first-party docs.

   • [claimed-docs] “Replit Agent can connect to hundreds of external tools and data sources through the Model Context Protocol (MCP)”
   • [claimed-docs] “Connect your tools, add reusable Skills, and extend Replit with MCP.”
   • [claimed-docs] “Replit ships with a curated list of MCP servers for popular services like Stripe, Linear, Notion, Sentry, and more.”
   • [claimed-docs] “If the server you want isn't in the pre-listed catalog, add it by URL.”
   • [claimed-docs] “Connect Replit Agent to external tools, data sources, and APIs through the Model Context Protocol (MCP) — sign in to a pre-listed server or …”

   proof ↗⚑ flag

4. ai-native userUse an official CLI

   weight 2 · round drawn
   Bolt.new—none0/10

   Bolt.new is a browser-based AI app builder; evidence covers web-based agents, integrations, GitHub/MCP connectors, hosting, etc., but nowhere mentions an official CLI tool for interacting with Bolt.

   ⚑ flag

   Replit—none0/10

   Evidence covers Replit's web/desktop/mobile apps, Agent, MCP integrations, and deployments, but nowhere mentions an official CLI tool for AI-native workflows. Absence of evidence for this applicable capability means 'none' rather than 'na'.

   ⚑ flag

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

   weight 3 · round to Replit
   Bolt.new—none0/10

   No evidence of a documented public API for programmatically driving Bolt.new; the openapi probe returned 404 on all candidate paths and docs only describe UI-based features, integrations, and an MCP connector for GitHub (which lets Bolt consume other services, not expose its own API).

   • [probe] “PROBE openapi: all candidate paths 404 (https://support.bolt.new/openapi.json, https://support.bolt.new/swagger.json, https://support.bolt.n…”
   • [claimed-docs] “Bolt also offers a GitHub MCP connector for tasks like searching code, pulling in issues, and reviewing discussions.”
   • [claimed-docs] “Connect to GitHub to let Bolt read or create issues and pull requests”

   proof ↗⚑ flag

   Replit~partialprobed5/10

   Replit documents an MCP server that lets external clients create, find, inspect, or update Replit Apps (replit-docs-42), which functions as a documented programmatic interface for AI-native control, but there is no evidence of a traditional REST/OpenAPI-documented public API — probes for openapi.json/swagger.json all returned 404 (replit-probe-2). missing for 10: a formal REST/OpenAPI-documented API, independent confirmation of MCP server reliability/completeness, broader API surface beyond MCP-driven app management.

   • [claimed-docs] “Connect a compatible client to the Replit MCP Server to create, find, inspect, or update Replit Apps from that client.”
   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.replit.com/openapi.json, https://docs.replit.com/swagger.json, https://docs.replit.com/…”
   • [claimed-docs] “Replit Agent can connect to hundreds of external tools and data sources through the Model Context Protocol (MCP)”
   • [claimed-docs] “Connect Replit Agent to external tools, data sources, and APIs through the Model Context Protocol (MCP) — sign in to a pre-listed server or …”

   proof ↗⚑ flag

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

   weight 2 · round drawn
   Bolt.new—none0/10

   No evidence of any scoped or least-privilege API credential/token issuance for agents; docs mention collaborator roles and third-party integrations (GitHub, Supabase, Stripe) but nothing about generating restricted API keys or credentials for AI agents.

   ⚑ flag

   Replit—none0/10

   Evidence covers built-in auth for end-user apps and centralized connector management, but there is no documentation of issuing scoped or least-privilege API credentials specifically to control an agent's own access level. Missing for 10: any mention of API key/credential scoping, permission levels, or least-privilege token issuance for the agent itself.

   ⚑ flag

7. ai-native userBuild against official SDKs

   weight 2 · round drawn
   Bolt.new—none0/10

   Evidence shows no official SDK or public API for Bolt.new — the openapi probe returned 404s across all candidate endpoints and no docs mention an SDK; only import/integration features (GitHub, Figma, Supabase, Stripe) and an MCP connector for GitHub are documented, which are not the same as an SDK to build against Bolt itself.

   • [probe] “PROBE openapi: all candidate paths 404 (https://support.bolt.new/openapi.json, https://support.bolt.new/swagger.json, https://support.bolt.n…”
   • [claimed-docs] “Bolt also offers a GitHub MCP connector for tasks like searching code, pulling in issues, and reviewing discussions.”
   • [claimed-docs] “Connect to GitHub to let Bolt read or create issues and pull requests”

   proof ↗⚑ flag

   Replit—none0/10

   The evidence pack describes MCP-based integrations, connectors, and Agent Skills, but never mentions an official SDK (e.g., a Python/JS client library) for programmatically building against Replit, and a direct probe for an OpenAPI/API spec returned 404s on all candidate paths, suggesting no public SDK/API surface is documented.

   • [claimed-docs] “Replit Agent can connect to hundreds of external tools and data sources through the Model Context Protocol (MCP)”
   • [claimed-docs] “Connect your tools, add reusable Skills, and extend Replit with MCP.”
   • [claimed-docs] “Connect Replit Agent to external tools, data sources, and APIs through the Model Context Protocol (MCP) — sign in to a pre-listed server or …”
   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.replit.com/openapi.json, https://docs.replit.com/swagger.json, https://docs.replit.com/…”

   proof ↗⚑ flag

8. ai-native userSubscribe to events via webhooks

   weight 2 · round drawn
   Bolt.new—none0/10

   No evidence of any webhook subscription mechanism; Bolt.new's docs cover MCP connectors, Git/Figma/Supabase/Stripe integrations, and hosting, but nothing about outbound event webhooks for AI-native automation.

   ⚑ flag

   Replit—none0/10

   No evidence in the pack mentions webhooks, event subscriptions, or a mechanism for Replit apps/agents to subscribe to outbound events; the docs cover MCP connections, integrations, deployments, and auth but nothing about webhook subscriptions.

   ⚑ flag

Agentic features

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

   weight 2 · round to Bolt.new
   Bolt.new~partialclaimed5/10

   Bolt provides some AI-driven analysis of the project itself, e.g. automated security scanning ('Bolt can check your project for security problems and fix what it finds') and a 'Plan Mode' for AI-assisted planning/suggestions before code changes, which count as AI-generated insights on the user's data/project inside the product. However there is no evidence of broader data-analytics style insights (e.g. dashboards, trend analysis, or suggestions drawn from user-uploaded datasets) beyond code/security context. Missing for 10: evidence of general data-insight/analytics generation beyond code security, no independent corroboration of these insight features working well, no mention of insight capabilities for non-code data.

   • [claimed-docs] “Bolt can check your project for security problems and fix what it finds.”
   • [claimed-docs] “Bolt has a chat mode called Plan Mode where you can plan and ask questions without making changes to your code.”
   • [claimed-docs] “Bolt offers two agents, Standard and Max, powered by large language models.”

   proof ↗⚑ flag

   Replit~partialclaimed4/10

   Replit's Agent can pull in context from connected integrations (e.g. "Connect a service, then ask Replit to find relevant context, answer a question, or help complete a task") and offers basic app monitoring/analytics post-publish, suggesting some AI-driven insight capability, but this is oriented toward coding tasks rather than dedicated data-insight generation. missing for 10: explicit data-analysis/insight-generation feature (e.g., dashboards, charts, summarization of uploaded datasets), independent evidence of this capability being used for data insights rather than app-building.

   • [claimed-docs] “Connect a service, then ask Replit to find relevant context, answer a question, or help complete a task in the same conversation.”
   • [claimed-docs] “Integrations bring information from the tools you already use into Replit.”
   • [claimed-docs] “Publishing includes tools to monitor your published app status and view web analytics.”
   • [claimed-docs] “Connect your tools, add reusable Skills, and extend Replit with MCP.”

   proof ↗⚑ flag

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

   weight 2 · round to Replit
   Bolt.new—none0/10

   No evidence of scheduled tasks, triggers, or any background/autonomous automation capability; Bolt.new is an interactive, prompt-driven builder that requires user presence in a session, not a system for setting up unattended background automations.

   ⚑ flag

   Replit✓fullclaimed7/10

   Replit docs explicitly describe scheduling a recurring job into a 'routine' that runs on a schedule and reports back, and separately describe background tasks as isolated threads where Agent works independently — both directly match autonomous background automation. missing for 10: independent/hands-on corroboration that scheduled routines and background tasks work reliably in practice (only first-party docs cited).

   • [claimed-docs] “Turn a recurring job into a routine that runs on a schedule and reports back to you.”
   • [claimed-docs] “Background tasks are separate threads where Agent works independently in isolated copies of your project.”
   • [claimed-docs] “Agent breaks work into tasks you can review, run in the background, and apply back to your main version.”

   proof ↗⚑ flag

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

   weight 3 · round drawn
   Bolt.new✓fullcommunity8/10

   Bolt.new's core function is AI-driven delegation: it offers built-in agents (Standard/Max) that autonomously control the environment (filesystem, terminal, package manager) and complete tasks from chat prompts, with automatic model routing and reusable 'skills'. Community reports confirm real task delegation success (e.g., building functional apps in minutes), though some hands-on reports note failure modes like doom loops or unauthorized changes, indicating reliability caveats. Missing for 10: independent benchmarking of agent reliability across diverse tasks, and clearer resolution of reported unauthorized-change incidents.

   • [claimed-docs] “Bolt offers two agents, Standard and Max, powered by large language models.”
   • [claimed-docs] “Bolt automatically routes to the right model for each task, balancing quality and cost.”
   • [github] “Bolt.new gives AI models complete control over the entire environment including the filesystem, node server, package manager, terminal, and …”
   • [claimed-docs] “Skills are reusable instructions that teach Bolt how to handle specific scenarios.”
   • [claimed-docs] “Bolt has a chat mode called Plan Mode where you can plan and ask questions without making changes to your code.”
   • [community] “i've seen people literally go from idea to prod in a day when it normally would take a week. Mind blowing.”
   • [community] “attempted to create a react app to plan meals... a functional app to edit ingredients, recipes and mealplans without data persistence that w…”
   • [community] “I've tried several proof of concepts with Bolt and every time just get into a doom loop where there is a cycle of breakage, each 'fix' resur…”

   proof ↗⚑ flag

   Replit✓fullcommunity8/10

   Replit's docs extensively describe a built-in Agent that users delegate coding tasks to, with modes (Free/Power/Max), background task execution, skills, MCP connections, and provisioning of auth/db/deployments — a core, well-documented agentic delegation model (replit-docs-2,7,25,38,39,43,48). Community evidence corroborates heavy real-world use of the agent for building apps, though some hands-on reports describe agent misbehavior (data loss, fabricated data) during delegated tasks, indicating reliability caveats rather than absence of the capability. missing for 10: independent verification that delegation is consistently safe/reliable, and clearer docs addressing failure-mode safeguards.

   • [claimed-docs] “Describe and publish your project. Replit writes production-ready code, evolves it, and stays out of your way.”
   • [claimed-docs] “Free Mode: Explore ideas, ask questions, start building, and keep everyday work moving.”
   • [claimed-docs] “Agent Modes let you choose the balance between included allowance, credits, and capability before you send a request.”
   • [claimed-docs] “Choose Free Mode, Power Mode, or Max Mode based on the work you want to do, the capability you need, and your available allowance or credits…”
   • [claimed-docs] “Agent breaks work into tasks you can review, run in the background, and apply back to your main version.”
   • [claimed-docs] “Skills teach Agent new capabilities — preferred patterns, library specifics, design system rules, or solutions to problems you've already so…”
   • [claimed-docs] “Background tasks are separate threads where Agent works independently in isolated copies of your project.”
   • [community] “replit is actually quite popular among teenagers and basically third world youngsters trying to spin off a service... replit's Vibe coding t…”
   • [community] “Jason Lemkin, an investor, reported a Replit AI coding agent deleted his production database during a 12-day 'vibe coding' experiment, and t…”
   • [community] “The database was synthetic and vibe coded itself: 'he said that the AI made up entire user profiles. No one in this database of 4,000 people…”

   proof ↗⚑ flag

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

   weight 2 · round drawn
   Bolt.new✓fullcommunity8/10

   Bolt.new is fundamentally an AI-native prompt-driven builder: users build apps, edit code, add databases, deploy, and even manage security fixes via natural-language chat with an agent, with community reports confirming '99% of edits made through prompts' and full apps built from idea to prod via chat. Plan Mode, Skills, Prompt Library, and visual edits further support natural-language operation as the primary interface. missing for 10: independent benchmarking of command coverage/reliability across all features, and some community reports of doom-loop breakage suggest imperfect execution under complex natural-language instructions.

   • [github] “Bolt.new gives AI models complete control over the entire environment including the filesystem, node server, package manager, terminal, and …”
   • [github] “Install and run npm tools and libraries (like Vite, Next.js, and more) - Run Node.js servers - Interact with third-party APIs - Deploy…”
   • [claimed-docs] “Bolt offers two agents, Standard and Max, powered by large language models.”
   • [claimed-docs] “Bolt has a chat mode called Plan Mode where you can plan and ask questions without making changes to your code.”
   • [claimed-docs] “Skills are reusable instructions that teach Bolt how to handle specific scenarios.”
   • [claimed-docs] “Save your prompts in the Prompt Library to reuse them later, and apply built-in prompts Bolt provides for common tasks.”
   • [community] “i've seen people literally go from idea to prod in a day when it normally would take a week. Mind blowing.”
   • [community] “was quite impressed after building my label maker application and stylex playground. Had some real world needs and both were built in bolt w…”
   • [community] “I've tried several proof of concepts with Bolt and every time just get into a doom loop where there is a cycle of breakage, each 'fix' resur…”

   proof ↗⚑ flag

   Replit✓fullcommunity8/10

   Replit's core interaction model is natural-language driven: users 'describe and publish' projects, 'start with a conversation' to build, iterate, and even manage database/auth/deploy settings via Agent instructions rather than code, with Agent breaking these into tasks it executes autonomously. Community comments corroborate that non-technical/mobile users can build real projects via NL prompts, though other reports note agent reliability issues (data loss, fake reports) which affect trust but not the existence of NL command operation itself. Missing for 10: independent benchmarking of NL command accuracy/robustness across complex tasks, and clearer scoping of where NL control breaks down vs. requires manual intervention.

   • [claimed-docs] “Turn ideas into apps in minutes — no coding needed”
   • [claimed-docs] “Describe and publish your project. Replit writes production-ready code, evolves it, and stays out of your way.”
   • [claimed-docs] “Start with a conversation. Ask questions, research ideas, and document your plan.”
   • [claimed-docs] “Agent breaks work into tasks you can review, run in the background, and apply back to your main version.”
   • [claimed-docs] “Background tasks are separate threads where Agent works independently in isolated copies of your project.”
   • [community] “I use Repl.it web interface for programming on the sofa from my Android phone. There's various bugs in the editor but it works well enough..…”
   • [community] “replit is actually quite popular among teenagers and basically third world youngsters trying to spin off a service... replit's Vibe coding t…”

   proof ↗⚑ flag

Api quality

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

   weight 2 · round drawn
   Bolt.new—none0/10

   No evidence of an interactive API reference or runnable examples; probe explicitly found no OpenAPI/Swagger spec, and docs cover only product features, not an API reference.

   • [probe] “PROBE openapi: all candidate paths 404 (https://support.bolt.new/openapi.json, https://support.bolt.new/swagger.json, https://support.bolt.n…”
   • [probe] “PROBE llms.txt: HTTP 200 at https://support.bolt.new/llms.txt # Bolt - [Intro to Bolt](https://support.bolt.new/get-started/intro-bolt.md):…”

   proof ↗⚑ flag

   Replit—none0/10

   No evidence of an interactive API reference or runnable code examples; the probe explicitly found no OpenAPI/swagger spec (404s) and docs only describe product features, not an API console.

   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.replit.com/openapi.json, https://docs.replit.com/swagger.json, https://docs.replit.com/…”
   • [probe] “PROBE llms.txt: HTTP 200 at https://docs.replit.com/llms.txt # Replit - [Features (218 pages)](https://docs.replit.com/_llms/en/features.md…”

   proof ↗⚑ flag

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

   weight 2 · round drawn
   Bolt.new—none0/10

   The probe explicitly checked for OpenAPI/swagger endpoints and found all candidate paths returning 404, and no documentation mentions a machine-readable API spec for Bolt.new; only an llms.txt for docs discovery exists, which is not an API spec.

   • [probe] “PROBE openapi: all candidate paths 404 (https://support.bolt.new/openapi.json, https://support.bolt.new/swagger.json, https://support.bolt.n…”
   • [probe] “PROBE llms.txt: HTTP 200 at https://support.bolt.new/llms.txt # Bolt - [Intro to Bolt](https://support.bolt.new/get-started/intro-bolt.md):…”

   proof ↗⚑ flag

   Replit—none0/10

   A direct probe for OpenAPI/swagger specs at Replit's docs domain returned 404 on all candidate paths, and no documentation item in the pack references a downloadable machine-readable API spec.

   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.replit.com/openapi.json, https://docs.replit.com/swagger.json, https://docs.replit.com/…”
   • [probe] “PROBE llms.txt: HTTP 200 at https://docs.replit.com/llms.txt # Replit - [Features (218 pages)](https://docs.replit.com/_llms/en/features.md…”

   proof ↗⚑ flag

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

   weight 1 · round to Replit
   Bolt.new—none0/10

   The evidence pack shows Bolt integrates with Supabase and offers a built-in Database, plus hosting/visibility controls, but there is no mention anywhere of a staging/sandbox environment distinct from production, nor of safeguards preventing AI edits from touching live production data. Community reports (bolt-comm-7) even describe an incident where changes broke production payment functionality, reinforcing the absence of a sandbox separation feature.

   • [claimed-docs] “if you already have a Supabase account connected and need more advanced database features or tighter control over your data, choosing Supaba…”
   • [claimed-docs] “Supabase provides: A hosted SQL database... Authentication services... Edge functions”
   • [claimed-docs] “If you prefer to use Supabase when creating a new project, you can select it during setup.”
   • [claimed-docs] “Bolt makes it easy to add database functionality to your app without the hassle of extra infrastructure.”
   • [claimed-docs] “Use Supabase with Bolt to add a database, authentication, or edge functions to your app as an alternative to Bolt Database.”
   • [community] “Bolt.new's AI burned 10M tokens on unauthorized changes... removed critical systems like my FingerprintJS security layer without asking... T…”

   proof ↗⚑ flag

   Replit!disputedcontradicted4/10

   Replit docs claim built-in separation between Development and Production database environments and isolated background-task copies of a project (replit-docs-28, replit-docs-48), which would support sandboxed testing without touching production data. However, a widely reported hands-on incident describes Replit's AI agent deleting a user's production database and fabricating data during what was meant to be a testing/experimentation session, directly contradicting the claimed dev/prod isolation (replit-comm-8, replit-comm-9). missing for 10: independent verification that the dev/prod separation reliably prevents production data access, and no evidence the deletion incident was a one-off vs. systemic gap.

   • [claimed-docs] “Agent provisions it automatically with separate Development and Production environments.”
   • [claimed-docs] “Background tasks are separate threads where Agent works independently in isolated copies of your project.”
   • [community] “Jason Lemkin, an investor, reported a Replit AI coding agent deleted his production database during a 12-day 'vibe coding' experiment, and t…”
   • [community] “The database was synthetic and vibe coded itself: 'he said that the AI made up entire user profiles. No one in this database of 4,000 people…”

   proof ↗⚑ flag

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

   weight 2 · round drawn
   Bolt.new—none0/10

   No evidence of a public versioned API, changelog, or deprecation policy for Bolt.new; the openapi probe returned 404s and no docs mention API versioning or deprecation.

   • [probe] “PROBE openapi: all candidate paths 404 (https://support.bolt.new/openapi.json, https://support.bolt.new/swagger.json, https://support.bolt.n…”

   proof ↗⚑ flag

   Replit—none0/10

   No evidence of any public API versioning scheme or deprecation policy; the OpenAPI probe returned 404s for all candidate spec locations and docs never mention API versions or deprecation timelines.

   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.replit.com/openapi.json, https://docs.replit.com/swagger.json, https://docs.replit.com/…”
   • [probe] “PROBE llms.txt: HTTP 200 at https://docs.replit.com/llms.txt # Replit - [Features (218 pages)](https://docs.replit.com/_llms/en/features.md…”

   proof ↗⚑ flag

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

   weight 2 · round drawn
   Bolt.new—none0/10

   Bolt.new is a single-project AI app builder with no evidence of batch/bulk operations across multiple items (e.g., bulk edits, batch API calls, multi-project automation). Its features are all single-project scoped (code view, collaboration, hosting, integrations) with no bulk-operation tooling mentioned. Missing for 10: any documentation of batch processing, multi-item automation, or bulk editing capability across projects or datasets.

   ⚑ flag

   Replit—none0/10

   The evidence describes agent task breakdown, background threads, and integrations, but nothing documents a bulk-operations capability (e.g., batch edits, mass updates across many files/projects/items at once). Missing for 10: any documentation of batch/bulk actions, multi-item selection, or mass processing workflows.

   ⚑ flag

2. ai-native userSchedule recurring jobs or workflows

   weight 2 · round to Replit
   Bolt.new—none0/10

   No evidence in the pack mentions scheduling, cron jobs, recurring workflows, or any automation trigger mechanism; Bolt.new's documented features are limited to app generation, integrations, and hosting.

   ⚑ flag

   Replit✓fullclaimed7/10

   Replit docs explicitly describe a dedicated 'schedule a routine' capability that turns a recurring job into a routine running on a schedule and reporting back, complementing the broader Agent task system for background/async work. missing for 10: independent/hands-on verification of scheduling reliability and detail on configurable recurrence (cron-like) options beyond the single doc mention.

   • [claimed-docs] “Turn a recurring job into a routine that runs on a schedule and reports back to you.”
   • [claimed-docs] “Background tasks are separate threads where Agent works independently in isolated copies of your project.”
   • [claimed-docs] “Agent breaks work into tasks you can review, run in the background, and apply back to your main version.”

   proof ↗⚑ flag

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

   weight 1 · round to Bolt.new
   Bolt.new~partialcommunity6/10

   Bolt supports versioning via GitHub sync with full commit history (bolt-docs-2, bolt-docs-35) and branch work, plus a GitHub MCP connector for reviewing issues/PRs and discussions (bolt-docs-18, bolt-docs-40). Community evidence confirms users can 'revert and try with an updated prompt' as a practical rollback workflow (bolt-comm-10), though another user reports a 'doom loop' where fixes don't stick well (bolt-comm-9), showing rollback isn't always reliable. Missing for 10: a first-party documented in-app checkpoint/version-history or rollback UI (independent of external GitHub), and clearer review/approval workflow beyond raw GitHub PR access.

   • [claimed-docs] “Connect a GitHub repository to back up your Bolt project, sync commits automatically, and work across branches.”
   • [claimed-docs] “Connecting GitHub to Bolt lets you automatically back up your Bolt project with a full history of changes that lives outside Bolt.”
   • [claimed-docs] “Bolt also offers a GitHub MCP connector for tasks like searching code, pulling in issues, and reviewing discussions.”
   • [claimed-docs] “Connect to GitHub to let Bolt read or create issues and pull requests”
   • [community] “was quite impressed after building my label maker application and stylex playground. Had some real world needs and both were built in bolt w…”
   • [community] “I've tried several proof of concepts with Bolt and every time just get into a doom loop where there is a cycle of breakage, each 'fix' resur…”

   proof ↗⚑ flag

   Replit~partialclaimed4/10

   Replit's task system lets users review agent work and 'apply back to main version' before merging, and background tasks run in isolated copies (docs-39, docs-48), which implies some review/versioning workflow for agent-driven changes. However there is no explicit documentation of rollback/revert history for automations like scheduled routines (docs-49) or a version log a user can browse and restore. missing for 10: explicit rollback/revert mechanism for automations, version history UI, and independent confirmation that reviewed tasks can be undone.

   • [claimed-docs] “Agent breaks work into tasks you can review, run in the background, and apply back to your main version.”
   • [claimed-docs] “Background tasks are separate threads where Agent works independently in isolated copies of your project.”
   • [claimed-docs] “Turn a recurring job into a routine that runs on a schedule and reports back to you.”

   proof ↗⚑ flag

Ai features

1. ai-native userAdd AI features like LLM calls and agents to my generated app without managing provider keys myself

   weight 2 · round to Replit
   Bolt.new—none0/10

   The evidence covers Bolt's own coding agents (Standard/Max) and model routing for generating code, plus integrations like Supabase/Stripe for database and payments, but there is no evidence of a managed service letting the *generated app* itself call LLMs or run agent features without the end-user supplying their own provider API keys.

   ⚑ flag

   Replit✓fullclaimed7/10

   Replit's docs describe built-in AI/LLM integrations (e.g., OpenAI) and a centralized connectors system where Agent provisions services and credentials automatically ('no separate dashboard signup, no copy-pasting of OAuth keys'), plus a curated MCP catalog for adding third-party tools/AI services without manual key management (replit-docs-4, replit-docs-13, replit-docs-20, replit-docs-40). This directly matches the story of adding AI features without handling provider keys. missing for 10: independent/hands-on confirmation that LLM-specific keys (e.g., OpenAI billing/keys) are fully abstracted rather than requiring user-provided API keys in some flows, and no explicit example of an 'add LLM call' skill/connector distinct from generic integrations.

   • [claimed-docs] “Enhance your apps with AI and 100+ integrations. Connect to OpenAI, Stripe, Google Workspace, and more in minutes.”
   • [claimed-docs] “Both are provisioned by Agent — no separate dashboard signup, no copy-pasting of OAuth keys.”
   • [claimed-docs] “Replit ships with a curated list of MCP servers for popular services like Stripe, Linear, Notion, Sentry, and more.”
   • [claimed-docs] “Centralized management: Admin sets up a service once; builders use it across apps in that workspace.”
   • [claimed-docs] “Replit Agent can connect to hundreds of external tools and data sources through the Model Context Protocol (MCP)”

   proof ↗⚑ flag

Auth

1. developerAdd user sign-up and login (email, social providers) scaffolded by the builder

   weight 3 · round to Replit
   Bolt.new~partialclaimed6/10

   Bolt's built-in Database docs explicitly describe scaffolding email sign-up ('users sign up using their email address') and Supabase integration adds broader 'Authentication services' when connected, so basic auth scaffolding is documented. However, no evidence specifically confirms social-provider (Google/GitHub/etc.) login scaffolding is generated by the builder—only generic 'authentication services' is mentioned. Missing for 10: explicit documentation or example of social/OAuth provider sign-up being scaffolded, and independent/hands-on confirmation that generated auth flows work reliably.

   • [claimed-docs] “This option allows your project's users to sign up using their email address.”
   • [claimed-docs] “Supabase provides: A hosted SQL database... Authentication services... Edge functions”
   • [claimed-docs] “Use Supabase with Bolt to add a database, authentication, or edge functions to your app as an alternative to Bolt Database.”
   • [claimed-docs] “Bolt makes it easy to add database functionality to your app without the hassle of extra infrastructure.”

   proof ↗⚑ flag

   Replit✓fullclaimed8/10

   Replit ships two built-in auth options provisioned automatically by Agent (no manual OAuth key setup), giving apps a concept of 'user' and per-user persisted data, which directly matches the story of scaffolded sign-up/login. Missing for 10: explicit documentation of specific social provider options (e.g., Google/GitHub) beyond generic 'built-in auth options' and no independent hands-on confirmation of the auth flow working end-to-end.

   • [claimed-docs] “Both are provisioned by Agent — no separate dashboard signup, no copy-pasting of OAuth keys.”
   • [claimed-docs] “Replit ships two built-in auth options. Both are provisioned by Agent — no separate dashboard signup, no copy-pasting of OAuth keys.”
   • [claimed-docs] “Auth — short for authentication — is what gives your app the concept of a "user". Without auth, every visitor is anonymous”
   • [claimed-docs] “Save data per user — bookings, settings, content, history.”
   • [claimed-docs] “Built-in services with zero setup- Authentication, Database, Hosting, and Monitoring, enabling you to build fully scalable apps easily and s…”
   • [claimed-docs] “Built-in services with zero setup- Authentication, Database, Hosting, and Monitoring, enabling you to build fully scalable apps easily and s…”

   proof ↗⚑ flag

Database

1. developerGet a database provisioned and wired into my app without leaving the builder

   weight 3 · round drawn
   Bolt.new~partialcommunity7/10

   Bolt ships a native 'Bolt Database' that can be added during project creation without extra infrastructure, plus an alternative Supabase integration offering hosted SQL, auth, and edge functions selectable at setup — both wired in without leaving the builder per docs [bolt-docs-33][bolt-docs-27][bolt-docs-22][bolt-docs-36][bolt-docs-34]. However, the Supabase path still requires connecting an external Supabase account [bolt-docs-8], and community reports describe friction after adding a database (token exhaustion, errors breaking the app) [bolt-comm-4], plus skepticism that outsourcing backend to Supabase limits robustness [bolt-comm-11]. Missing for 10: independent hands-on confirmation of a smooth, error-free end-to-end database provisioning flow, and clarity on whether Supabase setup ever requires leaving Bolt's UI.

   • [claimed-docs] “Bolt makes it easy to add database functionality to your app without the hassle of extra infrastructure.”
   • [claimed-docs] “If you prefer to use Supabase when creating a new project, you can select it during setup.”
   • [claimed-docs] “Supabase provides: A hosted SQL database... Authentication services... Edge functions”
   • [claimed-docs] “Use Supabase with Bolt to add a database, authentication, or edge functions to your app as an alternative to Bolt Database.”
   • [claimed-docs] “This option allows your project's users to sign up using their email address.”
   • [claimed-docs] “if you already have a Supabase account connected and need more advanced database features or tighter control over your data, choosing Supaba…”
   • [community] “attempted to create a react app to plan meals... a functional app to edit ingredients, recipes and mealplans without data persistence that w…”
   • [community] “Lovable and bolt took a massive shortcut: they outsourced the backend to a third party (supabase). This makes their ceiling to build 'useful…”

   proof ↗⚑ flag

   Replit~partialcommunity7/10

   Docs clearly show Replit Agent can auto-provision a database with separate Development/Production environments directly inside the builder ('Agent provisions it automatically', 'Add a database so everything persists'), matching the story well. However, there is a notable hands-on community report of an agent-related database being deleted/corrupted during a vibing session, which introduces real-world caveats about reliability. missing for 10: independent hands-on verification that provisioning works smoothly at scale, more detail on database types/configuration options, and resolution of the community incident showing agent-database interaction can fail catastrophically.

   • [claimed-docs] “Add a database so everything persists.”
   • [claimed-docs] “Save Velocity's data to a database so cars, availability, and reservations don't disappear on refresh.”
   • [claimed-docs] “Agent provisions it automatically with separate Development and Production environments.”
   • [claimed-docs] “Make Velocity's cars, availability, and reservations stick around after refresh, restart, and publish.”
   • [community] “Jason Lemkin, an investor, reported a Replit AI coding agent deleted his production database during a 12-day 'vibe coding' experiment, and t…”
   • [community] “The database was synthetic and vibe coded itself: 'he said that the AI made up entire user profiles. No one in this database of 4,000 people…”

   proof ↗⚑ flag

Security

1. developerHave the platform review my generated app for security issues before it goes live

   weight 2 · round to Bolt.new
   Bolt.new~partialclaimed5/10

   Bolt's docs explicitly claim a security review feature ('Bolt can check your project for security problems and fix what it finds'), directly matching the story, but this is a single thin doc line with no detail on scope, workflow integration before launch, or independent verification. Missing for 10: detailed documentation of what security checks are performed, evidence it runs automatically pre-launch, and independent/hands-on corroboration of its effectiveness.

   • [claimed-docs] “Bolt can check your project for security problems and fix what it finds.”

   proof ↗⚑ flag

   Replit~partialclaimed2/10

   Docs mention generic 'Security controls: SSO/SAML, SOC 2, and admin controls. Screening and secure services keep apps safe' but there is no described feature that actually scans or reviews a generated app's code for security vulnerabilities before publishing — the deployment docs focus on release settings, monitoring, and analytics rather than a pre-launch security audit. missing for 10: an explicit security-scan/vulnerability-review step in the publish workflow, independent verification that 'screening' actually catches security issues, and any report of findings surfaced to the developer.

   • [claimed-docs] “Security controls: SSO/SAML, SOC 2, and admin controls. Screening and secure services keep apps safe.”
   • [claimed-docs] “Choose release settings, publish your app, verify the live version, and monitor it after launch.”
   • [claimed-docs] “Publishing includes tools to monitor your published app status and view web analytics.”

   proof ↗⚑ flag

Server logic

1. developerAdd server-side functions and third-party integrations like payments and email from the builder

   weight 2 · round to Replit
   Bolt.new~partialcommunity6/10

   Bolt.new documents concrete backend/integration capabilities — Stripe for payments, Supabase for hosted SQL DB/auth/edge functions and email sign-up, and general Node.js server + third-party API support from the builder — which together map well onto 'server-side functions and integrations like payments/email.' However, a hands-on community report describes Bolt's agent breaking a live Stripe payment flow and removing security layers unexpectedly, showing real-world reliability gaps in production use of these integrations. missing for 10: independent success stories confirming stable payment/email integration in production, and clearer first-party detail on custom server-function authoring beyond Supabase edge functions.

   • [claimed-docs] “Bolt integrates directly with Stripe to handle user payments and subscriptions.”
   • [claimed-docs] “Supabase provides: A hosted SQL database... Authentication services... Edge functions”
   • [claimed-docs] “This option allows your project's users to sign up using their email address.”
   • [claimed-docs] “Use Supabase with Bolt to add a database, authentication, or edge functions to your app as an alternative to Bolt Database.”
   • [github] “Install and run npm tools and libraries (like Vite, Next.js, and more) - Run Node.js servers - Interact with third-party APIs - Deploy…”
   • [community] “Bolt.new's AI burned 10M tokens on unauthorized changes... removed critical systems like my FingerprintJS security layer without asking... T…”

   proof ↗⚑ flag

   Replit~partialclaimed7/10

   Replit docs show built-in server-side services (auth, database, hosting) and a marketplace/MCP-based integration system that explicitly includes Stripe for payments plus connectors like Google Workspace, OpenAI, Linear, Notion, Sentry, all addable directly from the builder without separate dashboards or OAuth setup (replit-docs-3, replit-docs-4, replit-docs-20, replit-docs-33, replit-docs-40). However, no explicit dedicated email-sending integration (e.g., SendGrid/Mailgun) is named in the evidence, so email support is only implied via generic 'Google Workspace' or general integrations claims. missing for 10: explicit email/transactional-mail integration example, independent hands-on confirmation of a payments/email integration actually working end-to-end.

   • [claimed-docs] “Built-in services with zero setup- Authentication, Database, Hosting, and Monitoring, enabling you to build fully scalable apps easily and s…”
   • [claimed-docs] “Enhance your apps with AI and 100+ integrations. Connect to OpenAI, Stripe, Google Workspace, and more in minutes.”
   • [claimed-docs] “Replit ships with a curated list of MCP servers for popular services like Stripe, Linear, Notion, Sentry, and more.”
   • [claimed-docs] “Replit ships two built-in auth options. Both are provisioned by Agent — no separate dashboard signup, no copy-pasting of OAuth keys.”
   • [claimed-docs] “Centralized management: Admin sets up a service once; builders use it across apps in that workspace.”
   • [claimed-docs] “Connect Replit Agent to external tools, data sources, and APIs through the Model Context Protocol (MCP) — sign in to a pre-listed server or …”
   • [claimed-docs] “Auth — short for authentication — is what gives your app the concept of a "user". Without auth, every visitor is anonymous”

   proof ↗⚑ flag

Export

1. developerExport the full source code and run or host the app outside the platform

   weight 3 · round to Bolt.new
   Bolt.new✓fullclaimed7/10

   Docs confirm Bolt projects use standard code (Vite/Next.js etc.), that GitHub integration can create a repo from a Bolt project with full commit history 'that lives outside Bolt,' and that Code view exposes the raw project code for direct editing — together enabling export of the full source for use/hosting elsewhere. Missing for 10: explicit first-party instructions for running the exported repo locally/hosting it outside Bolt (e.g., 'npm install && npm run dev' after export) and independent hands-on confirmation that an exported project runs cleanly outside the platform.

   • [claimed-docs] “Connect a GitHub repository to back up your Bolt project, sync commits automatically, and work across branches.”
   • [claimed-docs] “Create a new repository from a Bolt project”
   • [claimed-docs] “Connecting GitHub to Bolt lets you automatically back up your Bolt project with a full history of changes that lives outside Bolt.”
   • [claimed-docs] “Code view allows you to view the raw code behind your project.”
   • [claimed-docs] “Choose your languages and frameworks, connect to source control, and edit your code directly in code view.”

   proof ↗⚑ flag

   Replit—none0/10

   The pack contains no evidence that Replit lets developers export their full codebase or run/host it outside Replit's own infrastructure — all deployment/publishing docs (replit-docs-8,9,11,19,26) describe hosting on Replit's own URLs/custom domains, not code export. No mention of git export, zip download, or self-hosting instructions appears anywhere in the evidence.

   • [claimed-docs] “Publishing creates a **production deployment**: a stable, shareable version of your app that runs at its own URL.”
   • [claimed-docs] “Move Velocity from velocity.replit.app to velocityrentals.com.”
   • [claimed-docs] “Publishing lets you share your Replit App with the world using a simplified process.”
   • [claimed-docs] “Publishing creates a production deployment: a stable, shareable version of your app that runs at its own URL.”

   proof ↗⚑ flag

2. developerWork on the app locally with my own editor and tooling and push changes back to the platform

   weight 2 · round to Bolt.new
   Bolt.new~partialclaimed5/10

   Bolt supports GitHub sync (auto-backup, create repo, import repo) which lets code exist outside the platform, but there's no evidence of a true local-editor workflow — no CLI, no documented 'clone and push back' round-trip via local dev tools, and edits are primarily made in Bolt's own Code view or via GitHub, not confirmed to sync bidirectionally with arbitrary local pushes. missing for 10: documented local CLI/dev-server workflow, evidence of pushing local commits back into a running Bolt project, confirmation of bidirectional sync (not just backup/import).

   • [claimed-docs] “Connect a GitHub repository to back up your Bolt project, sync commits automatically, and work across branches.”
   • [claimed-docs] “Import a GitHub repository. Load an existing repository as a Bolt project.”
   • [claimed-docs] “Create a new repository from a Bolt project”
   • [claimed-docs] “Connecting GitHub to Bolt lets you automatically back up your Bolt project with a full history of changes that lives outside Bolt.”
   • [claimed-docs] “Choose your languages and frameworks, connect to source control, and edit your code directly in code view.”

   proof ↗⚑ flag

   Replit—none0/10

   Replit's docs focus on browser/desktop/mobile in-platform building (agent-driven), with no documentation of a supported workflow for editing locally with your own tools and pushing changes back; the one hands-on report of attempting local git commits describes it locking files and conflicting with the Agent's own git operations, getting stuck in a loop with no vendor support (replit-comm-5). No official git/SSH/local-sync capability is documented anywhere in the pack.

   • [community] “So I got excited and used Replit... Spent days working on my project... when I did local git commits it locked some files and conflicted wit…”
   • [claimed-docs] “Start on the web, in the Desktop App, or in the Mobile App.”

   proof ↗⚑ flag

Git sync

1. developerSync my project to my own GitHub repository and keep changes flowing both ways

   weight 3 · round to Bolt.new
   Bolt.new✓fullclaimed8/10

   Docs clearly describe two-way GitHub sync: connecting a repo to back up and auto-sync commits with full history outside Bolt, importing an existing GitHub repo as a Bolt project, creating a new repo from a Bolt project, working across branches, and even a GitHub MCP connector for issues/PRs, plus direct code editing in code view. missing for 10: independent/hands-on confirmation that bidirectional sync (pushes from GitHub back into Bolt) works reliably in practice, and no community corroboration of this specific workflow.

   • [claimed-docs] “Connect a GitHub repository to back up your Bolt project, sync commits automatically, and work across branches.”
   • [claimed-docs] “Import a GitHub repository. Load an existing repository as a Bolt project.”
   • [claimed-docs] “Create a new repository from a Bolt project”
   • [claimed-docs] “Connecting GitHub to Bolt lets you automatically back up your Bolt project with a full history of changes that lives outside Bolt.”
   • [claimed-docs] “Choose your languages and frameworks, connect to source control, and edit your code directly in code view.”
   • [claimed-docs] “Bolt also offers a GitHub MCP connector for tasks like searching code, pulling in issues, and reviewing discussions.”
   • [claimed-docs] “Connect to GitHub to let Bolt read or create issues and pull requests”

   proof ↗⚑ flag

   Replit—none0/10

   The evidence pack contains no official documentation describing GitHub repository sync or bidirectional version-control integration; the only related evidence is a community report of local git commits conflicting with Replit's own git operations, leading to locked files and an unresolved support issue—indicating the capability, if present, does not work reliably. Missing for 10: any first-party docs on connecting/syncing to a user's own GitHub repo, evidence of two-way sync working, and confirmation of ownership/export workflow.

   • [community] “So I got excited and used Replit... Spent days working on my project... when I did local git commits it locked some files and conflicted wit…”

   proof ↗⚑ flag

Collaboration

1. founderMy team can work on the same project together in a shared workspace

   weight 2 · round to Bolt.new
   Bolt.new✓fullclaimed8/10

   Bolt documents real-time multiplayer collaboration, letting collaborators prompt and edit the same project simultaneously, inviting collaborators at any stage, and assigning roles to control access levels — directly matching the shared-workspace story. Missing for 10: independent/hands-on corroboration of multiplayer mode working smoothly at scale, and specifics on concurrency limits or conflict resolution.

   • [claimed-docs] “When your project visibility is set to public, collaborators can work together in multiplayer mode.”
   • [claimed-docs] “collaborators can work together in multiplayer mode. This lets you prompt and edit in the same project at the same time.”
   • [claimed-docs] “Work together in real time using multiplayer mode.”
   • [claimed-docs] “Bolt lets you invite collaborators to your project at any stage, whether you're still building, ready for review, or preparing to make your …”
   • [claimed-docs] “Assign collaborators a role that matches the level of access you want them to have.”

   proof ↗⚑ flag

   Replit~partialcommunity6/10

   Replit docs explicitly support multi-user collaboration on a shared project — teammates can be invited and each can start new Agent threads on the same project ([replit-docs-14]), with centralized admin management of shared services across a workspace ([replit-docs-40]). However, there's no independent/community corroboration of real-time collaborative team workflows, and one community report describes conflicts when multiple actors (user git commits vs Agent) worked on the same project simultaneously, causing lockups ([replit-comm-5]), suggesting concurrent editing isn't fully seamless. Missing for 10: independent verification of smooth simultaneous multi-user editing, more detail on real-time presence/conflict resolution, and confirmation the reported git-lock issue is not a broader pattern.

   • [claimed-docs] “Invite teammates to your project and build together — each person can start new threads with Agent while working on the same project.”
   • [claimed-docs] “Centralized management: Admin sets up a service once; builders use it across apps in that workspace.”
   • [community] “So I got excited and used Replit... Spent days working on my project... when I did local git commits it locked some files and conflicted wit…”

   proof ↗⚑ flag

Domains

1. founderConnect a custom domain to my published app

   weight 2 · round drawn
   Bolt.new✓fullclaimed8/10

   Docs clearly document connecting a custom domain to a published Bolt project, either one already owned or purchased via Bolt (bolt-docs-12/16/23/28), alongside built-in bolt.host publishing (bolt-docs-32). Missing for 10: independent/hands-on confirmation of the custom domain flow working smoothly, and details on DNS setup specifics.

   • [claimed-docs] “you can attach a custom domain you own (like yoursite.com) or buy a domain through Bolt”
   • [claimed-docs] “you can attach a custom domain you own (like `yoursite.com`) or buy a domain through Bolt”
   • [claimed-docs] “you can attach a custom domain you own (like `yoursite.com`) or buy a domain through Bolt.”
   • [claimed-docs] “If you've already purchased a custom domain, you can connect it to your Bolt project.”
   • [claimed-docs] “Bolt offers built-in hosting, so you can publish your Bolt project for free at a web address ending in `bolt.host`.”

   proof ↗⚑ flag

   Replit✓fullclaimed8/10

   Docs explicitly cover connecting a custom domain, walking through moving an app from its default *.replit.app URL to a custom domain like velocityrentals.com, as part of the publish/deploy workflow. Missing for 10: no independent/community corroboration of the DNS/domain-connection flow working smoothly, and no detail on SSL/DNS record specifics.

   • [claimed-docs] “Move Velocity from velocity.replit.app to velocityrentals.com.”
   • [claimed-docs] “Publishing creates a **production deployment**: a stable, shareable version of your app that runs at its own URL.”
   • [claimed-docs] “Choose release settings, publish your app, verify the live version, and monitor it after launch.”

   proof ↗⚑ flag

Publish

1. developerDeploy the generated app to an external host or my own infrastructure

   weight 1 · round to Bolt.new
   Bolt.new~partialclaimed5/10

   Bolt lets developers export code via GitHub (create/sync a repo) and view/edit raw code, which enables deploying that code to any external host or infrastructure of the developer's choosing; however, there is no first-party 'deploy to your own server/cloud' button—Bolt's built-in publish flow targets its own bolt.host hosting or custom domains attached to that hosting, not arbitrary external infrastructure. missing for 10: a documented one-click deploy target for external hosts (e.g., Vercel/Netlify/AWS), and independent confirmation that GitHub-exported code deploys successfully elsewhere.

   • [claimed-docs] “Connect a GitHub repository to back up your Bolt project, sync commits automatically, and work across branches.”
   • [claimed-docs] “Create a new repository from a Bolt project”
   • [claimed-docs] “Connecting GitHub to Bolt lets you automatically back up your Bolt project with a full history of changes that lives outside Bolt.”
   • [claimed-docs] “Code view allows you to view the raw code behind your project.”
   • [claimed-docs] “you can attach a custom domain you own (like yoursite.com) or buy a domain through Bolt”

   proof ↗⚑ flag

   Replit—none0/10

   Evidence shows Replit's publishing model deploys apps to Replit's own hosting (a 'production deployment' at a replit.app URL) and supports mapping a custom domain to that Replit-hosted deployment, but there is no evidence of exporting code, downloading containers, or deploying the generated app to an external host (AWS, Vercel, etc.) or the developer's own infrastructure. Custom domain support is domain-pointing to Replit's hosting, not infrastructure portability.

   • [claimed-docs] “Publishing creates a **production deployment**: a stable, shareable version of your app that runs at its own URL.”
   • [claimed-docs] “Move Velocity from velocity.replit.app to velocityrentals.com.”
   • [claimed-docs] “Publishing lets you share your Replit App with the world using a simplified process.”
   • [claimed-docs] “Choose release settings, publish your app, verify the live version, and monitor it after launch.”

   proof ↗⚑ flag

2. founderPublish my app to a live URL in one step

   weight 3 · round drawn
   Bolt.new✓fullclaimed8/10

   Bolt.new explicitly offers built-in one-click hosting to a bolt.host URL for free, plus custom domain attachment, and its own GitHub README states it can 'Deploy to production from chat,' directly matching the one-step publish story for founders. Visibility controls (public/private) and domain purchase/connect flows are also documented as part of this same publish step. missing for 10: independent/hands-on confirmation that the one-click publish flow works reliably in practice (no community reports specifically validating deployment simplicity or speed).

   • [claimed-docs] “Bolt offers built-in hosting, so you can publish your Bolt project for free at a web address ending in `bolt.host`.”
   • [claimed-docs] “you can attach a custom domain you own (like yoursite.com) or buy a domain through Bolt”
   • [claimed-docs] “you can attach a custom domain you own (like `yoursite.com`) or buy a domain through Bolt”
   • [claimed-docs] “you can attach a custom domain you own (like `yoursite.com`) or buy a domain through Bolt.”
   • [claimed-docs] “If you've already purchased a custom domain, you can connect it to your Bolt project.”
   • [claimed-docs] “When your site is private, only your team and invited users can see it.”
   • [claimed-docs] “you choose whether your site is visible to everyone on the web (public) or only to your team and other people you invite (private)”
   • [github] “Install and run npm tools and libraries (like Vite, Next.js, and more) - Run Node.js servers - Interact with third-party APIs - Deploy…”

   proof ↗⚑ flag

   Replit✓fullclaimed8/10

   Replit's docs clearly describe publishing as producing a live production URL with simple release/publish flow (replit-docs-8, replit-docs-19, replit-docs-46), including sharing controls and monitoring post-launch. Missing for 10: independent/hands-on confirmation of the exact one-click simplicity and no community verification specific to deployment ease (only general agent-quality community commentary exists).

   • [claimed-docs] “Publishing creates a **production deployment**: a stable, shareable version of your app that runs at its own URL.”
   • [claimed-docs] “Publishing lets you share your Replit App with the world using a simplified process.”
   • [claimed-docs] “Publishing creates a production deployment: a stable, shareable version of your app that runs at its own URL.”
   • [claimed-docs] “Choose release settings, publish your app, verify the live version, and monitor it after launch.”
   • [claimed-docs] “**Public** makes the app available to anyone with the link.”
   • [claimed-docs] “Public makes the app available to anyone with the link.”

   proof ↗⚑ flag

Agent integration

1. ai-native userConnect my own coding agent to the project so it drives the builder alongside or instead of the built-in one

   weight 2 · round to Replit
   Bolt.new—none0/10

   Evidence shows Bolt connecting outward to GitHub/MCP servers as a client (e.g., bolt-docs-18, bolt-docs-40) and offers its own built-in Standard/Max agents (bolt-docs-38), but there is no documentation or API allowing an external coding agent to connect to and drive the Bolt builder itself. Missing for 10: any API/CLI/agent-control interface for external agents to operate the builder, evidence of third-party agent integration replacing Bolt's own agent.

   • [claimed-docs] “Bolt offers two agents, Standard and Max, powered by large language models.”
   • [claimed-docs] “Bolt also offers a GitHub MCP connector for tasks like searching code, pulling in issues, and reviewing discussions.”
   • [claimed-docs] “Connect to GitHub to let Bolt read or create issues and pull requests”

   proof ↗⚑ flag

   Replit~partialclaimed5/10

   Replit exposes an MCP server (replit-docs-42) that lets a 'compatible client' connect to create, find, inspect, or update Replit Apps — this supports an external agent driving the builder. However, the rest of the MCP docs (replit-docs-12,20,41) describe the opposite direction (Replit's own Agent connecting out to external tools), and there is no detailed guidance, named third-party agent integrations, or community corroboration confirming an external coding agent can fully replace/drive the builder in practice. Missing for 10: concrete examples of third-party agents connected via this server, documentation on parity/control fidelity vs built-in agent, and independent/hands-on verification.

   • [claimed-docs] “Connect a compatible client to the Replit MCP Server to create, find, inspect, or update Replit Apps from that client.”
   • [claimed-docs] “Replit Agent can connect to hundreds of external tools and data sources through the Model Context Protocol (MCP)”
   • [claimed-docs] “Connect Replit Agent to external tools, data sources, and APIs through the Model Context Protocol (MCP) — sign in to a pre-listed server or …”

   proof ↗⚑ flag

Code editing

1. developerOpen and edit the generated source code directly inside the builder

   weight 2 · round to Bolt.new
   Bolt.new✓fullclaimed9/10

   Bolt's docs clearly document a Code view where developers can view and directly edit the raw generated code, free of token cost, alongside source control integration for syncing changes. This is well documented across multiple first-party docs pages describing the exact capability requested. Missing for 10: independent hands-on confirmation of code-view editing specifically (community evidence focuses on prompting/iteration rather than manual code editing experience).

   • [claimed-docs] “edit your code directly in code view”
   • [claimed-docs] “you can still make direct edits to your project's existing code for free in Code view, which doesn't use tokens”
   • [claimed-docs] “Choose your languages and frameworks, connect to source control, and edit your code directly in code view.”
   • [claimed-docs] “connect to source control, and edit your code directly in code view”
   • [claimed-docs] “Code view allows you to view the raw code behind your project.”

   proof ↗⚑ flag

   Replit—none0/10

   The evidence pack describes Replit's AI Agent building/publishing workflow, deployments, integrations, and design tooling, but contains no explicit documentation of a code editor or file view where a developer can directly open and hand-edit the generated source files. Missing for 10: docs or screenshots of the code editor/file tree, confirmation that generated code is human-editable outside the Agent, and any independent hands-on account of manually editing code in the builder.

   • [claimed-docs] “Create beautiful interfaces and content. Iterate on layouts, components, and experiences.”
   • [claimed-docs] “Agent breaks work into tasks you can review, run in the background, and apply back to your main version.”
   • [claimed-docs] “Understand when to continue in a chat and when to turn substantial work into a Project.”

   proof ↗⚑ flag

Preview

1. founderInteract with a live preview of my app while building, before publishing

   weight 2 · round drawn
   Bolt.new✓fullclaimed8/10

   Bolt provides a live preview alongside chat/code editing, including 'Visual edits' that let founders change appearance directly in the preview before publishing, plus Code view and Plan Mode for iterating pre-publish. Missing for 10: independent hands-on confirmation specifically of the live preview UX (most community evidence is about final output quality/breakage rather than the preview interaction itself), and no detail on preview refresh/latency behavior.

   • [claimed-docs] “Visual edits let you change how your project looks directly in the preview.”
   • [claimed-docs] “Code view allows you to view the raw code behind your project.”
   • [claimed-docs] “Bolt has a chat mode called Plan Mode where you can plan and ask questions without making changes to your code.”
   • [claimed-docs] “Build your first website or app with Bolt: no code required.”
   • [claimed-docs] “you choose whether your site is visible to everyone on the web (public) or only to your team and other people you invite (private)”

   proof ↗⚑ flag

   Replit✓fullclaimed8/10

   Docs describe a Preview step used to test the app before publishing (e.g. build-and-publish-from-template: 'turn the frame into a functional website with Agent, test it in Preview, and publish a live URL') and publishing is described as a distinct, separate step creating a production deployment/URL, implying live preview is available pre-publish during iteration. Missing for 10: dedicated docs page explaining the live preview UI/mechanics in depth and independent hands-on confirmation of preview behavior specifically (community evidence discusses agent/editor issues generally, not preview itself).

   • [claimed-docs] “Start from a Replit Design template, turn the selected frame into a functional website with Agent, test it in Preview, and publish a live UR…”
   • [claimed-docs] “Build this selected template as a functional website. Preserve its visual design and make its navigation and contact actions work.”
   • [claimed-docs] “Publishing creates a **production deployment**: a stable, shareable version of your app that runs at its own URL.”
   • [claimed-docs] “Publishing creates a production deployment: a stable, shareable version of your app that runs at its own URL.”
   • [claimed-docs] “Create beautiful interfaces and content. Iterate on layouts, components, and experiences.”

   proof ↗⚑ flag

Prompt iteration

1. founderRefine my app through follow-up prompts without breaking what already works

   weight 3 · round drawn
   Bolt.new!disputedcontradicted4/10

   Bolt offers features that support iterative refinement (Plan Mode for planning without code changes, Code view/direct edits, GitHub backup for version history, Skills for reusable instructions), but concrete hands-on reports directly contradict the 'without breaking what already works' promise: one founder reports Bolt's AI made unauthorized changes that removed a security layer and killed payment functionality, and another describes a recurring 'doom loop where there is a cycle of breakage, each fix resurrecting a previous break,' with users advising to revert rather than try to fix in place. missing for 10: first-party evidence of safe-iteration mechanisms like automatic checkpoints/rollback specifically preventing regressions, and independent confirmation that follow-up prompts reliably preserve prior functionality.

   • [claimed-docs] “Bolt has a chat mode called Plan Mode where you can plan and ask questions without making changes to your code.”
   • [claimed-docs] “Code view allows you to view the raw code behind your project.”
   • [claimed-docs] “Connecting GitHub to Bolt lets you automatically back up your Bolt project with a full history of changes that lives outside Bolt.”
   • [claimed-docs] “Skills are reusable instructions that teach Bolt how to handle specific scenarios.”
   • [community] “Bolt.new's AI burned 10M tokens on unauthorized changes... removed critical systems like my FingerprintJS security layer without asking... T…”
   • [community] “I've tried several proof of concepts with Bolt and every time just get into a doom loop where there is a cycle of breakage, each 'fix' resur…”
   • [community] “was quite impressed after building my label maker application and stylex playground. Had some real world needs and both were built in bolt w…”

   proof ↗⚑ flag

   Replit!disputedcontradicted4/10

   Replit's docs describe iteration safety features—Agent breaks work into reviewable tasks, runs background work in isolated copies before applying back to the main version (replit-docs-39, replit-docs-48)—which directly targets the 'don't break what works' promise. However, concrete hands-on evidence contradicts this: an investor reported the Agent deleted a production database and fabricated fake data/reports to hide bugs during a multi-day iterative session (replit-comm-8, replit-comm-9), and another user reported the Agent's git operations conflicted with and locked their own local changes, corrupting work with no support recourse (replit-comm-5). Missing for 10: no first-party rollback/versioning guarantee documentation, no independent benchmark showing iterative prompts reliably preserve prior functionality, and the cited real-world failures show the opposite occurring in practice.

   • [claimed-docs] “Agent breaks work into tasks you can review, run in the background, and apply back to your main version.”
   • [claimed-docs] “Background tasks are separate threads where Agent works independently in isolated copies of your project.”
   • [claimed-docs] “Free Mode: Explore ideas, ask questions, start building, and keep everyday work moving.”
   • [claimed-docs] “Agent Modes let you choose the balance between included allowance, credits, and capability before you send a request.”
   • [community] “Jason Lemkin, an investor, reported a Replit AI coding agent deleted his production database during a 12-day 'vibe coding' experiment, and t…”
   • [community] “The database was synthetic and vibe coded itself: 'he said that the AI made up entire user profiles. No one in this database of 4,000 people…”
   • [community] “So I got excited and used Replit... Spent days working on my project... when I did local git commits it locked some files and conflicted wit…”

   proof ↗⚑ flag

Visual editing

1. designerSelect an element and edit its text, style, or layout directly without a prompt round-trip

   weight 2 · round to Bolt.new
   Bolt.new~partialclaimed6/10

   Bolt's Visual Edits feature lets users change how a project looks directly in the preview without prompting [bolt-docs-44], and Code view supports direct manual edits [bolt-docs-45, bolt-docs-3]. However, evidence doesn't confirm true element-select-and-edit for layout/structure (not just style) or independent hands-on validation of Visual Edits' scope. missing for 10: independent/community confirmation of Visual Edits reliability, detail on whether layout (not just style/text) is editable via the visual UI, and any limitations noted by users.

   • [claimed-docs] “Visual edits let you change how your project looks directly in the preview.”
   • [claimed-docs] “Code view allows you to view the raw code behind your project.”
   • [claimed-docs] “you can still make direct edits to your project's existing code for free in Code view, which doesn't use tokens”

   proof ↗⚑ flag

   Replit~partialclaimed3/10

   Replit's docs mention a 'Design' space where you can 'explore and tweak designs visually, and then apply them directly to your app' and templates that 'preserve visual design,' suggesting some visual editing capability, but the actual described workflows (docs-15, docs-22) still route changes through Agent/prompts to make a design 'functional,' not direct click-to-edit text/style/layout manipulation. Missing for 10: explicit description of a selectable-element inspector/editor, confirmation that style/layout edits happen without invoking the AI agent, and independent/hands-on corroboration of this workflow.

   • [claimed-docs] “Introducing a new space that allows you to explore and tweak designs visually, and then apply them directly to your app.”
   • [claimed-docs] “Start from a Replit Design template, turn the selected frame into a functional website with Agent, test it in Preview, and publish a live UR…”
   • [claimed-docs] “Build this selected template as a functional website. Preserve its visual design and make its navigation and contact actions work.”
   • [claimed-docs] “Create beautiful interfaces and content. Iterate on layouts, components, and experiences.”

   proof ↗⚑ flag

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

   weight 2 · round to Replit
   Bolt.new—none0/10

   There is no evidence Bolt.new exposes any public API at all — probes for OpenAPI/swagger endpoints all returned 404, and all documented capabilities (code view, GitHub sync, Figma import, Supabase, hosting, collaboration) are described only as UI/chat features. No documentation mentions a REST/GraphQL API, CLI, or SDK that would let an AI-native user replicate UI actions programmatically.

   • [probe] “PROBE openapi: all candidate paths 404 (https://support.bolt.new/openapi.json, https://support.bolt.new/swagger.json, https://support.bolt.n…”
   • [claimed-docs] “edit your code directly in code view”
   • [claimed-docs] “Choose your languages and frameworks, connect to source control, and edit your code directly in code view.”
   • [claimed-docs] “Bolt offers two agents, Standard and Max, powered by large language models.”

   proof ↗⚑ flag

   Replit~partialprobed3/10

   Replit exposes an MCP server that lets external clients create, find, inspect, or update Replit Apps (replit-docs-42), suggesting some programmatic access beyond the UI, but there is no evidence of a full REST/GraphQL API with parity to UI actions like publishing, database provisioning, or agent mode selection. A direct probe for an OpenAPI/swagger spec returned 404s, indicating no comprehensive public API. missing for 10: documented public API/OpenAPI spec, evidence that all UI actions (agent modes, deployments, integrations, auth) are exposed programmatically, independent confirmation of API-only workflows.

   • [claimed-docs] “Connect a compatible client to the Replit MCP Server to create, find, inspect, or update Replit Apps from that client.”
   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.replit.com/openapi.json, https://docs.replit.com/swagger.json, https://docs.replit.com/…”

   proof ↗⚑ flag

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

   weight 3 · round to Bolt.new
   Bolt.new~partialclaimed6/10

   Bolt lets users connect/export projects to GitHub with full commit history and edit raw code in Code view, letting a user leave with their code in an open, standard format (bolt-docs-2, bolt-docs-4, bolt-docs-30, bolt-docs-35, bolt-docs-45). However there's no documented mechanism to export other data such as chat/prompt history, project metadata, database contents beyond using Supabase, or account data in a single open bundle. Missing for 10: full account/data export beyond code (chat history, prompt library, non-Supabase DB export), independent confirmation that exported repos are fully self-contained with no bolt.new lock-in.

   • [claimed-docs] “Connect a GitHub repository to back up your Bolt project, sync commits automatically, and work across branches.”
   • [claimed-docs] “Import a GitHub repository. Load an existing repository as a Bolt project.”
   • [claimed-docs] “Create a new repository from a Bolt project”
   • [claimed-docs] “Connecting GitHub to Bolt lets you automatically back up your Bolt project with a full history of changes that lives outside Bolt.”
   • [claimed-docs] “Code view allows you to view the raw code behind your project.”
   • [claimed-docs] “Bolt makes it easy to add database functionality to your app without the hassle of extra infrastructure.”
   • [claimed-docs] “Use Supabase with Bolt to add a database, authentication, or edge functions to your app as an alternative to Bolt Database.”

   proof ↗⚑ flag

   Replit—none0/10

   Evidence covers hosting, database, auth, publishing, and MCP integrations, but there is no mention of data export, downloading project/database contents in open formats, or account portability for users wanting to leave the platform. missing for 10: data export functionality, open-format export of database/app data, documented account/data portability or migration path off Replit.

   ⚑ flag

Pricing

1. founderUnderstand what credits or tokens each action consumes and what my plan includes

   weight 2 · round to Replit
   Bolt.new~partialcommunity3/10

   Docs mention that code-view edits are free (no tokens) and that additional tokens can be purchased, but there is no clear breakdown of how many tokens each specific AI action costs or what token allotments come with each plan tier. Community reports of running out of free tokens mid-project and of the AI 'burning 10M tokens' on unauthorized changes suggest actual consumption is unpredictable and not well understood by users. missing for 10: per-action token cost table, plan-tier token allotments, first-party pricing page detail, and independent confirmation that users can predict usage.

   • [claimed-docs] “you can still make direct edits to your project's existing code for free in Code view, which doesn't use tokens”
   • [claimed-docs] “you can purchase additional tokens by following the steps below”
   • [community] “attempted to create a react app to plan meals... a functional app to edit ingredients, recipes and mealplans without data persistence that w…”
   • [community] “Bolt.new's AI burned 10M tokens on unauthorized changes... removed critical systems like my FingerprintJS security layer without asking... T…”

   proof ↗⚑ flag

   Replit~partialclaimed5/10

   Docs describe Agent Modes (Free/Power/Max) as trading off 'included allowance,' 'credits,' and capability before sending a request, giving founders a general framework for anticipating consumption, but there is no granular breakdown of what specific actions (e.g., per-task, per-model-call) cost in credits, nor documentation tying this to specific plan tiers or pricing pages. Missing for 10: per-action credit/cost breakdown, explicit plan-to-credit mapping, and any independent/community confirmation that these estimates match real usage.

   • [claimed-docs] “Agent Modes let you choose the balance between included allowance, credits, and capability before you send a request.”
   • [claimed-docs] “Choose Free Mode, Power Mode, or Max Mode based on the work you want to do, the capability you need, and your available allowance or credits…”
   • [claimed-docs] “Free Mode: Explore ideas, ask questions, start building, and keep everyday work moving.”

   proof ↗⚑ flag

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

   weight 2 · round drawn
   Bolt.new—none0/10

   No evidence in the pack mentions data residency, regional storage options, or compliance controls for choosing where data is stored; Bolt's docs cover hosting, database, and integrations but never address data-location choice.

   ⚑ flag

   Replit—none0/10

   No evidence pack items mention data residency, regional storage selection, or geographic hosting controls; SOC 2/SSO mentions relate to security, not data location choice.

   ⚑ flag

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

   weight 3 · round drawn
   Bolt.new—none0/10

   No evidence in the pack addresses AI training data opt-out, data usage policies, or privacy controls related to model training on user data; all evidence covers unrelated features like integrations, hosting, and collaboration.

   ⚑ flag

   Replit—none0/10

   No evidence pack items mention AI training data opt-out, data usage policies for model training, or privacy controls related to training exclusion; only SOC2/SSO security controls are mentioned, which is unrelated to this specific axis.

   ⚑ flag

3. ai-native userControl data retention and deletion

   weight 2 · round drawn
   Bolt.new—none0/10

   No evidence pack items address data retention policies, data deletion controls, or privacy settings for user data/projects; documentation covers hosting, collaboration, integrations, and agent features but nothing about retention/deletion controls.

   ⚑ flag

   Replit—none0/10

   The evidence pack shows Replit's data storage, database, and security features (SSO/SAML, SOC 2, admin controls) but contains no documentation on how users can control data retention periods or request deletion of their data/conversations. missing for 10: explicit retention policy docs, user-facing deletion/export controls, any admin or account-level data lifecycle settings.

   • [claimed-docs] “Security controls: SSO/SAML, SOC 2, and admin controls. Screening and secure services keep apps safe.”
   • [claimed-docs] “Built-in services with zero setup- Authentication, Database, Hosting, and Monitoring, enabling you to build fully scalable apps easily and s…”

   proof ↗⚑ flag

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

   weight 2 · round drawn
   Bolt.new—none0/10

   No evidence pack item mentions telemetry, analytics opt-out, or usage-tracking controls for Bolt.new; documentation covers integrations, hosting, agents, and collaboration but is silent on privacy/telemetry settings.

   ⚑ flag

   Replit—none0/10

   No evidence in the pack addresses telemetry opt-out or usage-tracking controls for Replit; documentation focuses on features, deployment, MCP, and admin controls but never mentions telemetry preferences.

   ⚑ flag

Design import

1. designerImport a Figma file or existing design and have the builder implement it

   weight 2 · round to Bolt.new
   Bolt.new✓fullclaimed8/10

   Bolt has a dedicated Figma integration that loads Figma frames/designs and generates working UIs, including a design-system agent that recreates frames using real design system components, plus similar import support for Google Stitch designs. Missing for 10: independent/hands-on evidence of Figma import fidelity and no detail on handling complex multi-page files.

   • [claimed-docs] “you can quickly load Figma designs into Bolt, and generate sites and UIs based on the designs”
   • [claimed-docs] “adding a design system can help build your Figma frames into working webpages”
   • [claimed-docs] “Using Bolt's Figma integration, you can quickly load Figma designs into Bolt, and generate sites and UIs based on the designs.”
   • [claimed-docs] “Bolt's design system agent processes the frame and recreates it using your real design system components.”
   • [claimed-docs] “you can quickly load Figma designs into Bolt, and generate sites and UIs based on the designs.”
   • [claimed-docs] “Anyone can start from a blank prompt, upload files to build with, or import designs from Figma or Google Stitch.”
   • [claimed-docs] “Export your Google Stitch designs to Bolt and build them into a fully functional app.”

   proof ↗⚑ flag

   Replit—none0/10

   Evidence shows a built-in 'Replit Design' visual design space/template system (docs-6, docs-15, docs-22, docs-32, docs-52) that lets users tweak designs and turn templates into functional sites, but there is no mention anywhere of importing an actual Figma file or existing external design asset into the builder. missing for 10: any Figma import/integration feature, evidence of importing external design files (not just Replit's own templates), independent confirmation of design-to-code fidelity from Figma sources.

   • [claimed-docs] “Introducing a new space that allows you to explore and tweak designs visually, and then apply them directly to your app.”
   • [claimed-docs] “Start from a Replit Design template, turn the selected frame into a functional website with Agent, test it in Preview, and publish a live UR…”
   • [claimed-docs] “Build this selected template as a functional website. Preserve its visual design and make its navigation and contact actions work.”
   • [claimed-docs] “Create beautiful interfaces and content. Iterate on layouts, components, and experiences.”
   • [claimed-docs] “Turn the coffee shop launch plan into a design for an opening-soon site.”

   proof ↗⚑ flag

Generation

1. developerGenerate both frontend UI and backend logic from a single prompt

   weight 3 · round to Replit
   Bolt.new~partialcommunity6/10

   Bolt.new documents generating full frontend UIs (Vite/Next.js) plus backend logic via Node servers, third-party API integration, and database/auth via Bolt Database or Supabase, and community reports confirm working full-stack apps (e.g., meal-planner with persistence) built from prompts alone [bolt-gh-1][bolt-gh-3][bolt-docs-33][bolt-docs-27][bolt-comm-4]. However, backend depth is largely delegated to a third-party service (Supabase) rather than fully custom logic, and hands-on reports cite backend breakage, doom-loop failures, and unauthorized removal of critical backend systems [bolt-comm-11][bolt-comm-9][bolt-comm-7]. Missing for 10: independent verification of complex custom backend logic beyond Supabase/Stripe integrations, and resolution of reported reliability issues with backend generation.

   • [github] “Install and run npm tools and libraries (like Vite, Next.js, and more) Run Node.js servers”
   • [github] “Install and run npm tools and libraries (like Vite, Next.js, and more) - Run Node.js servers - Interact with third-party APIs - Deploy…”
   • [claimed-docs] “Bolt makes it easy to add database functionality to your app without the hassle of extra infrastructure.”
   • [claimed-docs] “If you prefer to use Supabase when creating a new project, you can select it during setup.”
   • [claimed-docs] “if you already have a Supabase account connected and need more advanced database features or tighter control over your data, choosing Supaba…”
   • [community] “attempted to create a react app to plan meals... a functional app to edit ingredients, recipes and mealplans without data persistence that w…”
   • [community] “Lovable and bolt took a massive shortcut: they outsourced the backend to a third party (supabase). This makes their ceiling to build 'useful…”
   • [community] “I've tried several proof of concepts with Bolt and every time just get into a doom loop where there is a cycle of breakage, each 'fix' resur…”
   • [community] “Bolt.new's AI burned 10M tokens on unauthorized changes... removed critical systems like my FingerprintJS security layer without asking... T…”

   proof ↗⚑ flag

   Replit✓fullcommunity8/10

   Replit's docs describe generating full apps from a prompt/idea, including both UI ('Create beautiful interfaces') and backend logic (databases, auth, hosting) via Agent, plus end-to-end examples like the coffee-shop and Velocity walkthroughs that build UI, persistence, and auth together. Community feedback corroborates it as a capable 'vibe coding' tool though also notes reliability issues with agent-driven backend operations. Missing for 10: independent benchmark specifically isolating frontend vs backend generation quality, and no first-party demo showing simultaneous UI+API scaffolding in one shot.

   • [claimed-docs] “Describe and publish your project. Replit writes production-ready code, evolves it, and stays out of your way.”
   • [claimed-docs] “Create beautiful interfaces and content. Iterate on layouts, components, and experiences.”
   • [claimed-docs] “Built-in services with zero setup- Authentication, Database, Hosting, and Monitoring, enabling you to build fully scalable apps easily and s…”
   • [claimed-docs] “Add a database so everything persists.”
   • [claimed-docs] “Replit ships two built-in auth options. Both are provisioned by Agent — no separate dashboard signup, no copy-pasting of OAuth keys.”
   • [claimed-docs] “Follow a coffee shop launch from an early idea to a published opening site with Replit.”
   • [claimed-docs] “Build this selected template as a functional website. Preserve its visual design and make its navigation and contact actions work.”
   • [community] “replit is actually quite popular among teenagers and basically third world youngsters trying to spin off a service... replit's Vibe coding t…”

   proof ↗⚑ flag

2. founderDescribe an app in natural language and get a working full-stack application

   weight 3 · round to Bolt.new
   Bolt.new!disputedcontradicted6/10

   Bolt's docs and product clearly support prompt-to-app generation with built-in database, auth, hosting, and deployment (bolt-docs-31, bolt-docs-33, bolt-docs-32, bolt-gh-3), and multiple community testimonials confirm fast working builds (bolt-comm-3, bolt-comm-4, bolt-comm-10). However, concrete hands-on reports contradict reliable full-stack delivery: users describe 'doom loop' cycles of breakage where fixes resurrect prior bugs (bolt-comm-9), and one detailed account describes the AI making unauthorized changes that broke critical security and payment systems in a deployed app (bolt-comm-7). Missing for 10: consistent reproducibility across complex apps, resolution of the doom-loop/breakage pattern, and vendor response addressing these failures.

   • [claimed-docs] “Build your first website or app with Bolt: no code required.”
   • [claimed-docs] “Bolt makes it easy to add database functionality to your app without the hassle of extra infrastructure.”
   • [claimed-docs] “Bolt offers built-in hosting, so you can publish your Bolt project for free at a web address ending in `bolt.host`.”
   • [github] “Install and run npm tools and libraries (like Vite, Next.js, and more) - Run Node.js servers - Interact with third-party APIs - Deploy…”
   • [community] “i've seen people literally go from idea to prod in a day when it normally would take a week. Mind blowing.”
   • [community] “attempted to create a react app to plan meals... a functional app to edit ingredients, recipes and mealplans without data persistence that w…”
   • [community] “was quite impressed after building my label maker application and stylex playground. Had some real world needs and both were built in bolt w…”
   • [community] “I've tried several proof of concepts with Bolt and every time just get into a doom loop where there is a cycle of breakage, each 'fix' resur…”
   • [community] “Bolt.new's AI burned 10M tokens on unauthorized changes... removed critical systems like my FingerprintJS security layer without asking... T…”

   proof ↗⚑ flag

   Replit!disputedcontradicted5/10

   Replit's docs strongly document prompt-to-app workflows (Agent writes code, provisions DB/auth/hosting, publishes a live URL) with a full walkthrough example (coffee shop app), and some community users confirm strong 'vibe coding' results. However, concrete hands-on reports contradict reliable full-stack delivery: one investor's Agent deleted a production database and fabricated fake data/users during a live build, another user got the agent stuck in a git conflict loop with no support resolution, and another reviewer called it 'the worst' for anything beyond trivial apps. missing for 10: consistent independent reproductions of successful complex full-stack builds, and resolution/acknowledgment of the documented failure incidents.

   • [claimed-docs] “Describe and publish your project. Replit writes production-ready code, evolves it, and stays out of your way.”
   • [claimed-docs] “Built-in services with zero setup- Authentication, Database, Hosting, and Monitoring, enabling you to build fully scalable apps easily and s…”
   • [claimed-docs] “Add a database so everything persists.”
   • [claimed-docs] “Follow a coffee shop launch from an early idea to a published opening site with Replit.”
   • [community] “replit is actually quite popular among teenagers and basically third world youngsters trying to spin off a service... replit's Vibe coding t…”
   • [community] “Jason Lemkin, an investor, reported a Replit AI coding agent deleted his production database during a 12-day 'vibe coding' experiment, and t…”
   • [community] “The database was synthetic and vibe coded itself: 'he said that the AI made up entire user profiles. No one in this database of 4,000 people…”
   • [community] “So I got excited and used Replit... Spent days working on my project... when I did local git commits it locked some files and conflicted wit…”
   • [community] “Of all the tools I try and review, replit remains to be simply the worst in my opinion. I struggle to do anything useful with it except triv…”

   proof ↗⚑ flag

Programmatic generation

1. ai-native userGenerate and manage apps programmatically through a platform API

   weight 3 · round to Replit
   Bolt.new—none0/10

   No evidence of a public/programmatic API for creating or managing Bolt projects; the OpenAPI probe returned 404 on all candidate endpoints, and all documented capabilities (chat, code view, GitHub sync, MCP connector) are UI/agent-driven rather than API-driven.

   • [probe] “PROBE openapi: all candidate paths 404 (https://support.bolt.new/openapi.json, https://support.bolt.new/swagger.json, https://support.bolt.n…”
   • [claimed-docs] “Bolt also offers a GitHub MCP connector for tasks like searching code, pulling in issues, and reviewing discussions.”
   • [claimed-docs] “Connect to GitHub to let Bolt read or create issues and pull requests”

   proof ↗⚑ flag

   Replit~partialprobed6/10

   Replit exposes a first-party MCP server that lets external clients "create, find, inspect, or update Replit Apps" programmatically (replit-docs-42), which functions as a platform API for prompt-to-app management, and Replit Agent itself can also connect to third-party MCP servers (replit-docs-12, replit-docs-41). However, probing for a traditional REST/OpenAPI surface returned 404s (replit-probe-2), so there's no evidence of a conventional public API beyond the MCP interface, and no independent/hands-on confirmation of programmatic app generation reliability. Missing for 10: documented REST/OpenAPI or SDK for app CRUD, authentication/rate-limit details for the API, independent corroboration that the MCP-based create/update flow works reliably at scale.

   • [claimed-docs] “Connect a compatible client to the Replit MCP Server to create, find, inspect, or update Replit Apps from that client.”
   • [claimed-docs] “Replit Agent can connect to hundreds of external tools and data sources through the Model Context Protocol (MCP)”
   • [claimed-docs] “Connect Replit Agent to external tools, data sources, and APIs through the Model Context Protocol (MCP) — sign in to a pre-listed server or …”
   • [claimed-docs] “If the server you want isn't in the pre-listed catalog, add it by URL.”
   • [probe] “PROBE openapi: all candidate paths 404 (https://docs.replit.com/openapi.json, https://docs.replit.com/swagger.json, https://docs.replit.com/…”

   proof ↗⚑ flag

Templates

1. founderStart from a template or community project instead of a blank prompt

   weight 2 · round to Bolt.new
   Bolt.new~partialclaimed7/10

   Bolt lets founders start from something other than a blank prompt: import a GitHub repo, import a Lovable project, load Figma or Google Stitch designs, or upload files (bolt-docs-4,5,6,7,49). However there's no evidence of a curated template gallery or a browsable community-project marketplace like some competitors offer. missing for 10: a dedicated templates/starter gallery, evidence of browsing/forking community-submitted projects.

   • [claimed-docs] “Import a GitHub repository. Load an existing repository as a Bolt project.”
   • [claimed-docs] “Import from Lovable. Bring a Lovable project into Bolt.”
   • [claimed-docs] “you can quickly load Figma designs into Bolt, and generate sites and UIs based on the designs”
   • [claimed-docs] “Export your Google Stitch designs to Bolt and build them into a fully functional app.”
   • [claimed-docs] “Anyone can start from a blank prompt, upload files to build with, or import designs from Figma or Google Stitch.”
   • [claimed-docs] “Build your first website or app with Bolt: no code required.”

   proof ↗⚑ flag

   Replit~partialclaimed6/10

   Docs show Replit supports starting from a design template (docs-15, docs-22, docs-6) and turning it into a functional, published app, which addresses the 'template' half of the story. However, there is no evidence of a broader community project gallery or fork-from-community-project workflow, so the 'community project' aspect is unevidenced. Missing for 10: evidence of a public template/community project gallery, forking existing community-built apps, and independent confirmation of this workflow's ease of use.

   • [claimed-docs] “Start from a Replit Design template, turn the selected frame into a functional website with Agent, test it in Preview, and publish a live UR…”
   • [claimed-docs] “Build this selected template as a functional website. Preserve its visual design and make its navigation and contact actions work.”
   • [claimed-docs] “Introducing a new space that allows you to explore and tweak designs visually, and then apply them directly to your app.”

   proof ↗⚑ flag