Published Docs MCP Server

The ClickHelp Published Docs MCP Server lets your readers connect their own AI agents — Claude, Cursor, ChatGPT, Gemini, and others — directly to your published documentation. It implements the Model Context Protocol (MCP), an open standard for connecting AI assistants to external data sources and tools.

An agent connected to your portal answers strictly from your published content instead of guessing, working from a stale copy of your docs, or waiting for someone to paste chunks of documentation into the chat. When the answer is not in your documentation, the agent says so.

If you are looking for a way to let your own team's AI agents work inside the portal — editing topics and using unpublished projects — see the Authoring MCP Server instead.


Licensing

The Published Docs MCP Server is included in the ClickHelp platform at no extra cost.

What an Agent Can Access

The Published Docs MCP Server works with public publications only. An agent connected to it sees exactly what an anonymous visitor to your portal sees—nothing more—and can read that content but never change it.

Content Available to agents
Public publications Yes, read-only
Restricted publications No
Projects (unpublished source content) No
The Published Docs MCP Server does not support authentication. An agent cannot connect as a Power Reader, so content restricted to Power Readers stays out of reach. If your readers need agent access to restricted content, use the Authoring MCP Server with a dedicated Power Reader account instead.

For the full list of tools and the applicable rate limits, see Published Docs MCP Server Reference.

How It Works

  1. You enable the Published Docs MCP Server in portal settings and publish your documentation.
  2. A reader opens any topic in your public documentation, clicks Copy page, and uses the Connect with MCP block to copy either the endpoint URL or a ready-to-run command for a supported agent.
    Alternatively, you can provide your readers with the endpoint URL directly in your content.
  3. The reader asks the agent a question in natural language.
  4. The agent decides whether to query your documentation and which tools to invoke.
  5. The Published Docs MCP Server returns the matching content from your public publications, and the agent answers with it.

To turn the feature on, see Set Up Published Docs MCP Server. To let your readers know it is available, see Tell Your Readers About Published Docs MCP Server.

The Published Docs MCP Server complements the other ways of making your documentation AI-friendly: llms.txt tells agents what exists, Markdown URLs let them fetch a known page, and MCP lets them search and navigate on their own.