MCP server

Your forum is also reachable from AI coding agents and assistants over the Model Context Protocol (MCP). Connect an MCP client with an API key and it gets a toolset covering the same capabilities as the REST API: reading and moderating questions, managing categories and tags, managing your team, and more, generated straight from the same underlying API, never a hand-written subset.

Endpoint

One URL works for every forum:

https://dashboard.forumcat.com/api/mcp

There is nothing to substitute: your API key (or approved connection) already identifies your forum, so the URL is the same for everyone and you can paste it as-is.

The same endpoint also answers at /api/mcp on your portal subdomain and on your custom domain, if you’d rather hand an agent a URL under your own brand. All three reach the same server with the same toolset; the URL above is simply the standard one to reach for.

The server speaks Streamable HTTP (the current MCP transport) and only accepts POST requests: it’s stateless, so there’s no session to open with a GET or close with a DELETE, and a client that tries either (including a regular browser visit) gets a plain 405.

Authentication

The MCP server uses the same API keys as the REST API, sent the same way:

Authorization: Bearer <your-api-key>

There’s no separate credential to create. A request with no key or an unrecognized one is rejected before the server does anything else.

What a key can do here

A key’s role and read-only modifier shape its toolset exactly the way they shape the REST API:

• Owner and Admin keys get the full management toolset: content, categories, tags, team, settings, and (owner/admin only) custom domains. Good for a general-purpose assistant you trust to run your forum.
• Agent keys get day-to-day content tools, answering and moderating, without categories, tags, team, settings, or custom domains. Good for a support bot that should only ever touch content.
• Read-only keys, layered on any role, only see read tools when the client asks what’s available. Good for an assistant that should look things up and never change anything, a research or reporting use case.

When your client asks the server what tools are available, it only lists the ones your key’s role can reach, and a tool that needs a specific role says so right in its description (for example, (requires role: owner, admin)). This is a convenience for discovery, not the actual security boundary: if a client somehow calls a tool outside what its key allows anyway, the call fails with the same error the REST API would give, an ordinary tool-level failure, not a crash or a silent no-op.

Managing a custom domain from an agent

An owner or admin key can connect, check the status of, and disconnect a forum’s custom domain the same way you’d do it from the dashboard’s Custom Domain page. Handy for scripting domain setup as part of onboarding a new forum, or for an agent that manages several forums’ domains on your behalf.

Connecting a client

Claude Code

claude mcp add --transport http myforum \
https://dashboard.forumcat.com/api/mcp \
--header "Authorization: Bearer <your-api-key>"

A generic JSON client config

Most MCP clients that support Streamable HTTP and custom headers use a config shape close to this:

{
"mcpServers": {
  "myforum": {
    "type": "http",
    "url": "https://dashboard.forumcat.com/api/mcp",
    "headers": {
      "Authorization": "Bearer <your-api-key>"
    }
  }
}
}

Check your specific client’s docs for the exact field names, they vary, but the two pieces (the Streamable HTTP URL and an Authorization header) are the same everywhere.

Clients without custom header support

Some MCP clients can’t attach custom headers to a Streamable HTTP connection. If yours can’t, a local bridge like mcp-remote can inject the header for you, or check whether a newer version of your client has added header support before working around it.

claude.ai custom connectors and Claude Desktop

claude.ai’s custom connectors and Claude Desktop’s connectors don’t let you attach a bearer header at all, so they can’t use an API key directly. Instead they connect with OAuth:

1. In claude.ai (or Claude Desktop), add a custom connector and paste the MCP URL: https://dashboard.forumcat.com/api/mcp.
2. The client discovers the connection automatically (standard OAuth discovery, no setup needed on your side) and opens your dashboard to sign in, if you aren’t already. This step registers the client with us automatically too (no account or setup needed on your side); it’s rate limited to a generous number of registrations per network per hour, so a normal connect never comes close to it.
3. You’ll see a screen asking you to approve the connection: which forum it’s for, and how much access to give it (the same Admin/Agent roles and read-only modifier an API key can have, capped by your own role). Approve it and the client connects, exactly like an API key would.

Only a forum owner or admin can approve a connection, the same as who can create an API key. The client never sees or stores an API key at all; it gets its own short-lived credentials instead, refreshed automatically in the background for as long as you keep using it.

Managing connected agents

Every app you’ve approved this way shows up on your dashboard’s API keys page, under Connected agents: which app, which forum, what role and read-only setting it was approved with, when it was approved, and when it was last used. Click Disconnect to revoke one immediately; the app loses access right away and would need to be reconnected and re-approved to use your forum again. Approving and disconnecting are both recorded in your forum’s audit log, the same as creating and revoking an API key.

Next steps

• API keys
• REST API