Business logic, ADRs, conventions — in Markdown, right in the project repo, next to your code. One command turns them into a docs site your team can browse and your AI can build on.
curl -fsSL https://raw.githubusercontent.com/documan-ai/documan/main/scripts/quickstart.sh | bash
Process
From Markdown to a live docs site in minutes
Docs live next to your code. ADRs, guides, specs — plain files, versioned in Git, reviewed in PRs.
One command. Docs site with semantic search is live. No npm, no config, no maintenance.
Your team browses the site. AI tools query docs via MCP. One source of truth for everyone.
Use Cases
Perfect for teams and solo devs alike
Your codebase changes every day. Your Confluence doesn't. Documan lives in your repo — every PR can update the docs, every merge keeps them in sync. No more "this is outdated, ask Jana."
Half the knowledge is in Slack threads nobody can find. The other half left with the last developer. Documan gives your team one place that's always current, searchable, and doesn't require asking someone.
Why does the pricing engine round down on the third decimal? Why do refunds go through a separate queue? That logic lives in someone's head — until they leave. Write it once, find it forever.
Stop spending your Fridays explaining how the system works to someone who joined Monday. New devs search, read, and ask questions — and get answers from your actual docs, not hallucinations.
Broken links and missing files don't announce themselves. They just quietly make your docs useless. Documan catches them in your pipeline before they reach anyone.
Your AI assistant that actually knows your project. Ask it why a decision was made, how a system works, or what the onboarding steps are — and get answers cited from your own documentation.
Features
Everything Docusaurus, VitePress, and Confluence can't do.
Your AI agent can search your docs semantically. Works with Claude Code, Cursor, and any MCP-compatible tool.
Find docs by meaning, not just keywords. Vector embeddings understand context and intent.
Docs are stored in your repository. Versioned, reviewed, and always in sync with your code.
No npm. No Node.js. No node_modules. Just one binary — install it once and never think about it again.
Lint docs in your pipeline. Broken links, missing frontmatter, and invalid references never reach production.
Add the Documan skill to your AI tool. It creates pages, manages navigation, and validates everything for you.
Feature-by-feature, no marketing fluff
| Feature | Us | Docusaurus |
|---|---|---|
| AI agentic web chat | ✓ BYOK |
✗ |
| Parallel research agents | ✓ | ✗ |
| Semantic search | ✓ BYOK |
✗ |
| MCP server | ✓ BYOK |
✗ |
| Agent config in ENV | ✓ | ✗ |
| Configurable AI endpoint | ✓ | ✗ |
| AI writing skill | ✓ | ✗ |
| Full-text search | ✓ | ✓ |
| Missing file detection | ✓ | ✗ |
| Frontmatter validation | ✓ | ✗ |
| Broken link detection | ✓ | ✓ |
| CI/CD linting | ✓ | ⚠ manual |
| ...and 14 more features | ||
| Feature | Documan | Docusaurus | VitePress | GitBook | Notion | Confluence |
|---|---|---|---|---|---|---|
| AI agentic web chat | ✓ BYOK | ✗ | ✗ | ✓ paid | ⚠ basic | ⚠ basic |
| Parallel research agents | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ |
| Semantic search | ✓ BYOK | ✗ | ✗ | ✓ paid | ⚠ basic | ⚠ basic |
| MCP server | ✓ BYOK | ✗ | ✗ | ✗ | ✗ | ✗ |
| Agent config in ENV | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ |
| Configurable AI endpoint | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ |
| AI writing skill | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ |
| Full-text search | ✓ | ✓ | ⚠ plugin | ✓ | ✓ | ✓ |
| Missing file detection | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ |
| Frontmatter validation | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ |
| Broken link detection | ✓ | ✓ | ⚠ plugin | ✗ | ✗ | ✗ |
| CI/CD linting | ✓ | ⚠ manual | ⚠ manual | ✗ | ✗ | ✗ |
| ...and 14 more features | ||||||
BYOK = Bring Your Own Key — your OpenAI API key for semantic search and the MCP server, your Anthropic API key for the agentic web chat
Pricing
Free forever for small projects
Get started for free and explore what Documan can do for your workflow
Unlimited files and the complete AI toolkit
Everything in Pro — one payment, no subscription
* BYOK (Bring Your Own Key) — your OpenAI API key powers semantic search and the MCP server; your Anthropic API key powers the agentic web chat.
Your developer's hour costs €50–100. Setting up existing solutions takes hours, not minutes — Documan pays for itself before lunch on day one.
FAQ
Confluence and Notion are collaboration tools, not developer tools. They live outside your repo, docs get out of sync, and AI agents can't read them. Documan keeps Markdown next to your code — versioned in Git, reviewed in PRs, readable by AI natively.
Yes, and you should. Markdown in your repo means AI reads it natively. The MCP server adds semantic search on top — it finds relevant snippets without loading everything into the context window. Both work, together they're strongest.
Absolutely — that is what the AI agentic web chat is for. Anyone can ask a question in plain, natural language, in their own language, and the agent finds the answer for them across your documentation or knowledge base — reading the relevant pages and replying with citations back to the source. No need to know the structure or hunt through pages by hand: readers, support teams, and non-technical users get a direct, grounded answer in the language they asked in.
Docusaurus and VitePress require Node.js, npm, plugin configs, and every upgrade risks breaking something. Documan is a single binary with zero dependencies — setup takes minutes, not hours. Semantic search, AI chat, MCP server, CI/CD linting — all included. Install once, never think about it again.
The free tier supports up to 100 files. Need more? Pick Pro or Unlimited and the limit is gone. If the pricing doesn't fit, write us — tell us what you need and what works for you. We're not here to gatekeep.
Your source files always stay in your Git repo — that never changes. Documan imports them into a database that can run self-hosted or managed by us — your choice. AI features use your own OpenAI or Anthropic API key, so your data stays yours. Need private or on-premise AI? No enterprise plan needed — the OpenAI and Anthropic endpoints are configurable via ENV, so you can route them through a private gateway or compatible endpoint and keep inference in your own cloud: AWS Bedrock, Google Vertex AI, Microsoft Foundry, or your own self-hosted models.
The curl script runs an install wizard that sets up everything. Got existing Markdown without frontmatter? ./documan fix handles it. Tested on a million files — fix, import, and live docs site in under 30 seconds.
Claude Code, Cursor, Windsurf, VS Code, Zed, Cline, GitHub Copilot, JetBrains AI, ChatGPT, and Claude Desktop — plus any MCP-compatible client. With OpenAI, Google, and Microsoft all backing it, MCP has become the de facto standard for AI tool integration, and the list keeps growing.
Your docs are plain Markdown files in your Git repo — always yours. Documan is just a renderer and search layer. The binary is versioned on GitHub and Docker Hub. Even if we disappeared — which we definitely don't plan — you download the binary and keep running.
Jiri Zapletal — software engineer, 21 years across SaaS, enterprise systems, and developer tools. One person so far, shipping fast. Public roadmap, issues on GitHub, full transparency.
Docusaurus is free. Your time isn't. Setting it up takes hours, adding search requires plugins, and every upgrade cycle costs more. Documan is running in 10 minutes with search, AI chat, MCP, and CI/CD linting included. At €7/month for your whole team, it pays for itself the first afternoon you don't spend debugging node_modules.
Business logic, ADRs, conventions — written once, in your repo forever. Anyone can read them. Any AI can build on them.