Documentation

The Analytics Power Hour Librarian lets an AI assistant search and quote the podcast's episode transcripts. It's a remote MCP (Model Context Protocol) server with two read-only tools.

Connect

Server URL (Streamable HTTP, no authentication):

https://mcp.analyticshour.io/mcp

Claude

On claude.ai or in Claude Desktop, open Customize → Connectors.
Select +, then Add custom connector.
Enter a name, such as "APH Librarian", and the server URL. Leave the OAuth fields in Advanced settings empty.
In a chat, turn the connector on from the tools menu, then ask a question about the show.

On Claude Team and Enterprise plans, an organization owner adds the connector first, under Organization settings → Connectors.

ChatGPT

Open Plugins, go to the MCPs tab, and choose Add → Add MCP server.
Set the type to Streamable HTTP, enter the server URL, and leave authentication empty.
Start a new thread, then ask a question about the show.

Other MCP clients

Any client that supports remote MCP servers over Streamable HTTP can connect with the URL above. The server supports MCP protocol versions 2025-03-26 through 2026-07-28 and needs no credentials.

Tools

`search_transcripts`

Finds passages about a topic, person, or idea. Each result includes the episode number, title, link, and release date, and each snippet line starts with its timestamp and speaker.

Parameter	Description
`query`	Required. What to look for, in natural language.
`episode_number`	Optional. Only search this episode, for example `"300"`.
`top_k`	Optional. Number of snippets to return, 1–20. Defaults to 8.

`list_episodes`

Lists episodes newest first, with each episode's number, title, release date, and link. It also returns the total number of episode transcripts (bonus episodes included) and the latest episode.

Parameter	Description
`year`	Optional. Only episodes released in this year, for example `"2024"`.
`title_contains`	Optional. Case-insensitive text that must appear in the title.
`limit`	Optional. Maximum episodes to return, 1–400. Defaults to 25.

Example questions

What does Jacob Matson say about semantic layers?
How many episodes are there, and what's the latest?
Which episodes came out in 2024?
Have the hosts talked about A/B testing pitfalls? Give me examples from a few episodes.

How it works

Coverage: every episode transcript in the show's archive is indexed in full. New and updated transcripts are picked up within about an hour.
Read-only: the tools only read the archive. Nothing you ask can change it.
Transcripts: quotes come from episode transcripts, which can contain transcription errors. Check the episode at the quoted timestamp before relying on exact wording.
Limits: requests are rate-limited per network address to keep the service available for everyone. If you see a rate-limit error, wait a minute and try again.

What the server receives and keeps is described in the privacy policy. For help, see support.