PodHood is in private beta — request access.
PodHoodDocs

MCP server

Every channel is an anonymous, read-only MCP server for AI agents.

Every channel on PodHood exposes an MCP server — anonymous, read-only, over Streamable HTTP:

https://<slug>.podhood.com/mcp

On a channel with a verified custom domain, the same server answers at https://<your-domain>/mcp.

There's no key to issue and nothing to configure — reading published content is anonymous by design (see auth.md). An Authorization header is accepted but ignored.

Your audience doesn't need this page, either: the Library's Connect AI button ships copy-paste install recipes for dozens of clients — Claude, ChatGPT, Cursor, VS Code, Zed, Gemini CLI, and more.

Add it to your client

claude mcp add --transport http yourshow https://yourshow.podhood.com/mcp

Add an HTTP MCP server pointing at the channel endpoint — no auth fields needed:

{
  "mcpServers": {
    "yourshow": {
      "url": "https://yourshow.podhood.com/mcp"
    }
  }
}

The generic Streamable-HTTP shape any MCP client accepts:

{
  "type": "http",
  "url": "https://yourshow.podhood.com/mcp"
}

Then ask anything — the agent answers from the podcast's own transcripts, with links to the second.

The search tool

The server exposes one tool, search: natural-language search over the channel's published episodes and transcripts. It returns ranked episodes and grounded, timestamped quotes, each with a citation URL to the exact moment.

Prop

Type

Protocol notes

  • Transport is stateless Streamable HTTP with JSON responses: JSON-RPC over POST; GET returns 405. CORS is open, so browser-based clients work.
  • Discovery surfaces: each host serves /.well-known/mcp (with a server card) and lists its MCP endpoint in llms.txt.
Was this page helpful?

On this page