Reference and guides drift apart
The API changes, the guides don’t, and readers find out the hard way.
For developers
API docs that fit your Git workflow.
Import an OpenAPI file for endpoint pages readers can try, write guides in Markdown synced with GitHub, and give AI tools clean text. Use the free cloud beta or your own servers.
Sound familiar?
Docs sit outside code review
Writers edit in one tool, engineers in another, and nobody reviews the same change.
AI coding tools read your docs badly
Assistants scrape your HTML pages instead of getting clean text they can use.
API reference
Reference pages from your OpenAPI file.
Import an OpenAPI 3.x file and Nibleaf builds endpoint and schema pages, with a try-it panel readers use in the browser.
- Public external references in your spec are fetched and bundled on import
- Write guides alongside the reference with code groups and request and response examples
- The Signal theme is laid out for code-heavy docs
Markdown and Git
Keep docs in your review process.
Connect a GitHub repository: browser edits land on a dedicated branch and pull request, and changes made in the repo flow back into Nibleaf.
- Pages are stored as Markdown you can read and diff
- Each pull request gets its own preview link, hidden from search engines
- When a page changes in both places, you resolve the conflict in Nibleaf
AI tools
Docs AI tools can actually read.
Every site serves llms.txt and llms-full.txt for AI tools. Private pages stay out of them.
- Connect Claude Code, Cursor or VS Code to your docs project over MCP
- Read only, Support writer and Full admin presets keep access scoped
- Publishing from an assistant always needs a fresh preview and confirmation
Requirements: Self-hosters turn it on first; it is off by default. Available on Nibleaf Cloud, and when you self-host v0.2.1 or later.
Self-hosting and versions
Run it where your code runs.
The whole product is open source under AGPL-3.0, with a Docker Compose installer for your own servers, and separate docs for each API version.
- Needs PostgreSQL, a Redis-compatible cache and S3-compatible storage; storage comes bundled
- Qdrant for Ask AI and ClickHouse for insights are optional
- Ask AI answers from the version and language the reader is viewing
How developer teams set it up
Import your spec
Upload an OpenAPI file, point to its URL, or pick it from your connected repository.
Connect GitHub
Link the repository so guide changes go through pull requests. Re-import the spec from it when the API changes.
Write the guides
Use the editor or your code editor, with code groups and request and response examples.
Publish and version
Preview, publish, and keep a separate version of the docs for each API release.
Questions from developers
Which OpenAPI versions are supported?
OpenAPI 3.x. Nibleaf validates the file and bundles public external references when you import it.
Does try-it work with any API?
Requests go from the reader’s browser to your API, so your API has to allow them (CORS) and readers supply their own credentials. Nibleaf doesn’t mock or monitor your API.
Can I write docs in my own code editor?
Yes. With GitHub two-way sync you edit Markdown in the repository and open pull requests as usual. Public GitLab and other Git repositories can be imported one way.
What do AI tools get from my docs?
Each site serves llms.txt and llms-full.txt, limited to content readers are allowed to see. Your team can also connect AI assistants to the project over MCP.
What does self-hosting need?
Docker Compose with PostgreSQL, a Redis-compatible cache and S3-compatible storage (bundled by default). Qdrant and ClickHouse are optional. The published image targets linux/amd64.
Ship API docs from your own workflow.
Start free on Nibleaf Cloud, or run it on your own servers.