Write documentation the way you write software
Markdown and MDX in your own editor, branches and pull requests, preview builds on every PR, API references generated from your OpenAPI spec, and an MCP server your coding agent can already read.
The world's best product teams trust Documentation.AI
From next-gen start-ups to established enterprises.
Why documentation loses to the sprint
Not because engineers do not care. Because the docs sit somewhere the workflow never reaches.
- The context switch tax
- Different tool, different login, different editor, none of your keybindings. A five-minute correction costs twenty.
- No review surface
- Content changes ship with no diff, no preview, and no way for anyone to approve them the way they approve code.
- The reference drifts from the spec
- The API changed in the handler and in the OpenAPI file. The page describing it still documents last quarter’s response shape.
- Coding agents cannot read your docs
- Cursor and Claude Code answer from stale training data, and every confident wrong guess becomes a support ticket.
Markdown, branches, pull requests, previews
Your Git repository is the source of truth. Documentation.AI builds a preview for every pull request and deploys when it merges.
- Write .mdx in your own editor, with 100+ components and your configuration in documentation.json
- Branch, commit, and open a pull request. Every PR gets its own preview build
- Merge to the deployment branch and Documentation.AI runs the production build
- Connect the Authoring MCP server so Cursor or Claude Code can edit pages, manage branches, and deploy
- Generate the API reference from your OpenAPI spec, playground included
- Automate the rest: broken link audits, API sync from code, and merge-driven documentation updates
Role-based permissions apply to the MCP server as well, so an agent gets exactly the read or write access its key allows and nothing more.
Built for the way you already work
Three pieces of the platform exist specifically because a developer would refuse to use it otherwise.
Two MCP servers, not one
The Reader MCP server gives AI tools grounded, read-only search over your published docs, including private and gated sites. The Authoring MCP server lets agents inspect files, update pages, manage branches, and deploy through your normal workflow.
API reference from the spec
Point a navigation group at an OpenAPI file and get parameter documentation, request examples in cURL, Python, JavaScript, Go, and Ruby, response examples for every status code, and a live playground.
Docs that maintain themselves
Workflows trigger on pull request merges. API sync from code reviews changes affecting the reference, and documentation update from code proposes edits wherever behaviour moved.
The parts that matter to an engineer
Nothing here asks you to leave your editor or learn a proprietary format.
Every pull request builds its own preview URL before anything reaches production.
Standard Markdown and MDX in your repository, edited wherever you already write code.
Read-only for your customers’ AI tools, read and write for your team’s.
Lighthouse Performance, Accessibility, and SEO with no build configuration to tune.
Frequently asked questions
The implementation questions, answered without a sales call.
Which Git providers are supported?
GitHub and GitLab. Repository-triggered workflows require a documentation site synced to a connected repository on one of them.
Can I keep my own CI?
Documentation.AI builds and deploys from your deployment branch and builds a preview for each pull request. Your existing checks still run on the same PR alongside it.
Does the MCP server work with private documentation?
Yes. Public sites use the /_mcp endpoint. Access-controlled sites use /_mcp/auth, which supports OAuth 2.1 for interactive clients and machine-to-machine credentials for automation.
What stops an agent commit from retriggering the workflow that made it?
Background workflows detect bot commits, so agent-authored changes do not retrigger the run that produced them.
Ready to build documentation people and AI agents love?
Live in under 5 min · No credit card required