The HelpDocs MCP (Model Context Protocol) server lets AI assistants like Claude and ChatGPT connect directly to your Knowledge Base. Once connected, your assistant can search, read, and even draft articles for you, without leaving the comfort of your AI assistant's interface.
MCP Capabilities
After connecting, your AI assistant will be able to:
- Search your articles and categories
- Read the full content of any article
- Create, update, archive, and delete articles and categories (including drafts and working copies)
- Add images to articles (not available in ChatGPT)
- Manage your help center's redirects
- Pull analytics, if your plan includes them
Your assistant can only do what your role and permission groups in HelpDocs allow, so someone who can't publish or change settings won't see those actions, and it can only reach the articles you can.
Connecting to MCP
First, enable MCP in HelpDocs:
- Head to Settings > AI
- Under For AI apps, turn on MCP access
- Hit Save changes
Once it's on, the same section shows your MCP server URL and quick setup steps for Claude, ChatGPT, Claude Code, Cursor, and VS Code. For a full walkthrough, see Claude, ChatGPT, Cursor, or VS Code.
The exact steps vary depending on which AI client you're using, but the setup process is similar across platforms. You'll need to provide your client with a name:
HelpDocs
and an MCP URL:
https://••••••••••••••••••••••••••
Once added, we'll walk you through the steps to Authorize MCP usage.
Connecting Without a Browser
The authorization flow above needs someone to click Authorize in a browser. If your client runs headless—a CI pipeline, a scheduled script, a self-hosted agent—there's nobody there to do that.
In that case you can authenticate with an API key instead, using a key you create in Settings > API.
Confirming It Works
Once connected, ask your assistant something like "Search my HelpDocs articles for billing". If it returns a list of your articles, you're all set ✅
Not seeing any results? Try this 👇
- Check MCP access is still on in Settings > AI, and that the URL matches the MCP server URL shown there
- If you connected with an API key, check the key still exists in Settings > API
- Make sure the host and path were entered exactly, with no trailing spaces
- Reconnect the client to refresh the authorization
That's it! Your AI assistant is now wired straight into your Knowledge Base 🚀