Vessel is an open-source, self-hostable documentation layer for the teams building the next interface. It turns a Git-backed source into:
- human docs that are readable and fast;
- llms.txt and llms-full.txt context for AI systems;
- an MCP-ready surface for discoverable resources and tools;
- a generated agent manifest, permission contract, and
skill.mdworkflow; - repo-native agent instructions and runnable examples;
- a maintenance layer that detects drift, mines support gaps, and routes explainable changes through review;
- a provider-neutral AI Assistant with BYOK support for OpenRouter, OpenAI, Anthropic, and custom OpenAI-compatible endpoints.
This repository contains the product landing page, a Markdown source compiler, config-driven navigation/version metadata, a generated agent contract, grounded search, a read-only docs audit, read-only MCP tools, and a static-first deployment scaffold. The interactive previews are intentionally local and dependency-light while the authenticated control plane, analytics, and maintenance services are being designed in public.
See FEATURES.md for the Mintlify-class feature inventory, current demo coverage, and the implementation order for the open-source product.
These screenshots are captured from the current local build. Open the visual tour for the full gallery, or start with the landing page below.
git clone https://github.com/everyai-com/mintly-alternative.git
cd mintly-alternative
npm install
npm run setup
npm run devOpen http://localhost:5173. npm run setup creates a local .env from .env.example without overwriting an existing file or adding secrets. See GETTING_STARTED.md for provider setup, deployment, and troubleshooting.
Useful commands:
npm run doctor # check Node, source files, generated artifacts, and config
npm run verify # build the site and run the complete test gate
npm start # alias for npm run devnpm install
npm run dev
Build and preview the production bundle:
npm run build
npm run preview
The Cloudflare Worker also exposes GET /health after deployment.
The functional docs surface is available at GET /docs/, GET /api/docs/index, GET /api/docs/search?q=checkout, GET /api/docs/page?slug=quickstart, GET /api/docs/audit, GET /api/ for the generated API reference and safe request preview, GET /api/openapi/index, GET /api/openapi/operation?id=createCheckoutSession, GET /agent-manifest.json, and POST /api/mcp. Netlify exposes the same contracts through its Functions redirects.
Put OpenAPI JSON, YAML, or YML contracts in openapi/. npm run docs:build turns them into static operation pages, JSON artifacts, MCP resources, and request examples. The playground builds curl/fetch snippets and previews documented responses; it deliberately does not proxy arbitrary authenticated requests from a public docs site.
netlify.toml is already configured with the Vite build command, dist output directory, SPA fallback, Netlify Functions, template environment prompts, and safe response headers.
The Netlify flow connects the new site to the GitHub repository and prompts for the optional assistant environment variables. GETTING_STARTED.md is included as the short post-deploy setup guide; USAGE.md contains the complete environment and API reference.
wrangler.jsonc uses a Worker entrypoint and static assets binding, including /api/assistant and /api/github/install:
For a local authenticated deploy:
npm run deploy
Cloudflare's Deploy to Cloudflare flow creates a GitHub/GitLab copy, configures the Worker, and sets up Workers Builds. Add the provider key after deployment with wrangler secret put AI_API_KEY; never commit it.
The browser includes a polished provider settings surface for OpenRouter, OpenAI, Anthropic, and custom OpenAI-compatible endpoints. It keeps a key in memory only for the current tab and calls the same-origin /api/assistant gateway. The gateway now grounds the provider prompt with the indexed Markdown pages and returns source paths alongside the answer. The production path is server-side configuration through Worker secrets or Netlify environment variables.
Supported variables are documented in .env.example. Provider-specific behavior is implemented in src/assistant-core.js; the Cloudflare Worker and Netlify Function share the same adapter logic.
OpenRouter is a useful default because it exposes an OpenAI-compatible chat API and a broad model catalog. You can still point Vessel directly at OpenAI, Anthropic’s native Messages API, or a trusted compatible endpoint.
The source repository is everyai-com/mintly-alternative. One-click deployment keeps GitHub in the loop for future pushes and previews. The optional GitHub App installer route is documented in GITHUB.md and intentionally stays disabled until GITHUB_APP_SLUG is configured.
docs.config.json is the portable site contract for navigation, versions, locales, agent capabilities, constraints, and freshness policy. npm run docs:check catches broken internal links, missing metadata, stale pages, missing examples, and navigation drift before a deployment. The generated agent-manifest.json, agent-permissions.json, and skill.md give coding agents an explicit read-only workflow instead of relying on page scraping or model memory.
The public page is deliberately framework-light for fast, portable self-hosting. shadcn/ui is a good open-source choice for the next authenticated editor/control plane because it distributes editable component source under MIT; see SHADCN.md for the decision and boundary.
The first version is being shaped around three recurring needs in agent-facing documentation:
- Agents need a compact map of a docs site, not only rendered HTML.
- Agents need explicit constraints, examples, and permission boundaries beside the implementation.
- Teams need an inspectable, self-hostable deployment path instead of a mandatory hosted control plane.
The product roadmap is:
- browser editing and MDX components on top of the current Markdown compiler and config-driven navigation/version contract;
- built-in components, themes, custom domains, and browser editing;
- OpenAPI/AsyncAPI reference rendering; the current OpenAPI JSON/YAML path already generates reference pages, examples, MCP operations, and a safe request-preview playground;
- generated llms.txt, full-context exports, agent contracts, quality audits, and an Assistant with richer citations and analytics;
- content drift detection, scheduled semantic audits, support-ticket adapters, and suggested reviewable PRs;
- a hybrid Git/visual editor with bidirectional sync, comments, previews, and GraphQL/style linting;
- public docs, customer help center, private wiki, embedded help widget, versions, localization, screenshots, and migration-safe redirects;
- authentication, page-level access, roles, approval stages, audit logs, and self-hosted analytics;
- changelogs, RSS, webhooks, Slack, support integrations, admin APIs, and MCP auth/streaming;
- an agent workflow that researches, plans, validates, and opens reviewable PRs.
The attached Mintlify-alternatives inventory informed this order. We selected the features that improve accuracy, ownership, and self-hosting first; the UI preview is not a claim that every listed backend exists today. See FEATURES.md for the category-by-category status and research links.
The product shape is informed by the current agent-doc primitives:
- the llms.txt proposal describes a concise Markdown overview for websites that is easier for LLMs to use than a full HTML crawl: https://llmstxt.org/
- the MCP specification defines discoverable tools and resources for model-controlled workflows: https://modelcontextprotocol.io/specification/2025-06-18/server/tools
- Netlify documents an official one-click Deploy button for public Git repositories: https://docs.netlify.com/deploy/create-deploys/
- Cloudflare documents Deploy to Cloudflare buttons for public repositories and automatic Workers setup: https://developers.cloudflare.com/workers/platform/deploy-buttons/
- shadcn/ui documents its open-code Vite setup: https://ui.shadcn.com/docs/installation/vite
- shadcn/ui’s source repository is MIT licensed: https://github.com/shadcn-ui/ui/blob/main/LICENSE.md
- OpenRouter documents its OpenAI-compatible quickstart: https://openrouter.ai/docs/quickstart
- OpenAI documents API authentication: https://developers.openai.com/api/reference/overview#authentication
- Anthropic documents its Messages API: https://platform.claude.com/docs/en/api/overview
- Mintlify's current platform surface is documented through its pricing feature matrix: https://www.mintlify.com/pricing
- Mintlify documents navigation, tabs, versions, and localization: https://mintlify.com/docs/navigation
- Mintlify documents OpenAPI/AsyncAPI interactive playgrounds: https://mintlify.com/docs/api-playground/overview
- Mintlify documents its component system: https://www.mintlify.com/mintlify/docs/components/index
- Mintlify documents Assistant citations and agentic retrieval: https://mintlify.com/docs/guides/assistant
- Mintlify documents search, feedback, and analytics: https://mintlify.com/docs/guides/analytics
- Mintlify documents the research-to-PR Agent workflow: https://www.mintlify.com/docs/agent
- GitBook documents Git sync, Assistant, Agent, and MCP: https://gitbook.com/docs
- GitBook documents extending Assistant with external MCP servers: https://gitbook.com/docs/publishing-documentation/gitbook-assistant
- Fern documents its documentation MCP server: https://buildwithfern.com/learn/docs/ai-features/mcp-server
- ReadMe documents API reference landing pages with authentication-aware setup: https://docs.readme.com/main/docs/reference-core-pages
- Docusaurus documents versioning: https://docusaurus.io/docs/versioning
- Docusaurus documents Git-based localization: https://docusaurus.io/docs/i18n/tutorial
MIT. See LICENSE.
Open-source infrastructure for local-first, governed AI agents:
- distillory — local-first memory engine that reasons at ingestion
- agentprofile — one agent identity — skills, credentials, memory — across every tool
- agent-ready — turn any app into an MCP server + API + CLI, safe by default
- primer — live business context injected into any agent
- plainsync — local-first Markdown workspace for humans + agents
- argus — cloud-native software verification with parallel browser testing
- talltrack — sales calls in, publishable content out
Built by Phanindra Reddy · magicteams.ai