AI-ready documentation

Documentation your AI tools can actually use.

Every Doccupine site ships an MCP server, so Claude, Cursor, and your own agents answer straight from your live docs, with citations, always current.

Included on every plan. No per-message billing.

Static docs weren't built for the AI era.

Your documentation is where the answers live. The problem is the tools your users reach for can't read it.

Users ask AI first.

People open Claude or ChatGPT before they open your docs. If those tools can't read your documentation, they answer from stale training data, or not at all.

LLMs hallucinate your API.

Without a live source to read from, models invent endpoints, parameters, and options that were never in your API. Confident, plausible, and wrong.

Search returns pages, not answers.

Full-text search hands back ten links and leaves the reading to the developer. What they wanted was one synthesized answer they can act on.

How it works

Connect your docs to any MCP client.

MCP is the open standard AI clients use to read external sources. Doccupine speaks it out of the box, so wiring your docs into Claude, Cursor, or your own agent is three steps.

  • Publish your Doccupine site.Every published site exposes an MCP server at /api/mcp automatically. Nothing to configure, nothing extra to deploy.
  • Add the server to Claude, Cursor, or your agent.Drop one entry into your mcpServers config and point it at your docs URL.
  • Ask in natural language.The client reads straight from your live docs and answers the question, citing the page it came from.
claude_desktop_config.json
{
  "mcpServers": {
    "your-docs": {
      "url": "https://docs.yoursite.com/api/mcp"
    }
  }
}

Your docs, as a graph your agents can traverse.

Each page is a node, wired to the pages it relates to. Your AI clients follow those links to find the exact answer, then cite where it came from.

Current, cited, and yours.

An answerable docs server is only worth it if its answers stay right, show their work, and respect your access rules.

Answers that stay accurate.

Pages are source-linked and drift-synced, so when your code changes the docs behind the MCP server track it. Answers keep up with your latest release.

Cited, not guessed.

Every answer points back to the exact page it came from, so your team can verify it and your users can read further.

Locked down when you need it.

Gate the server with an optional DOCS_API_KEY bearer token, so private documentation stays private while public docs stay open.

No extra infrastructure.

The MCP server ships with every published site, on every plan. No separate service to run and no per-message bill for answers.

Put it to work.

One MCP server, many places your docs can start answering for you.

Deflect support tickets.

Point your support assistant at your docs so it answers product questions from the source, with a link to the page, before they reach your inbox.

Answers inside the IDE.

Developers ask Cursor or Claude about your SDK without leaving the editor, and get answers drawn from your current reference docs.

Power your own agents.

Give your internal agents a reliable, cited source of product knowledge over MCP instead of pasting docs into prompts by hand.

Make your docs answerable.

Publish with Doccupine and every site ships an MCP server your AI tools can read. Free for 30 days, no credit card.