Podcast Index MCP Server & CLI

Podcast Index MCP Server & CLI

Podcast Index MCP server and CLI for Claude Code, Codex and AI agents. 36 tools for podcast search, episode transcripts you can actually read and search, Podcasting 2.0 tags, value-for-value data, feed health and index stats.

One install gives you both surfaces, the same 36 tools under the same names, reading one array of tool definitions so they cannot drift apart.

It searches four million podcasts, and then it does the thing the API stops short of: it opens the transcript. Podcast Index hands out a link to an episode's transcript file and goes no further, which is useless to an assistant that cannot click. This fetches the file, works out whether it is SRT, WebVTT, JSON or HTML, and gives you the actual words with timestamps and speakers.

So you can ask when something was said, and get an answer.

There are 36 tools. One free key covers all but two of them.

Built by Navid Moazzez. Built on Slipway, which turns one definition of each tool into the MCP server and the CLI.

Two ways to use it

Command line

podcastindex-cli runs every tool as a command. Agents that run commands, like Claude Code, Codex and OpenCode, use it on their own, and you can type the same commands in a terminal, a script or a cron job:

podcastindex-cli                                          # every command, one line each
podcastindex-cli status                                   # what this server can reach
podcastindex-cli search-podcasts --q "ai agents" --max 5
podcastindex-cli get-trending --max 10
podcastindex-cli search-transcript --episode-id <id> --query "compound interest"
podcastindex-cli find-guest-appearances --name "Tim Ferriss"
podcastindex-cli check-feed-health --show https://example.com/feed.xml
podcastindex-cli submit-feed --feed-url https://example.com/feed.xml --confirm
podcastindex-cli <command> --help                         # what any command takes

--confirm is the shell spelling of the confirmation that adding a feed to the index needs. --json gives JSON, --compact puts it on one line, --select keeps only the fields you name, and --agent is compact JSON with no prompts, and never confirms a write. Exit codes are 0 ok, 1 an unexpected error, 2 usage or a hidden or refused write, 3 not found, 4 auth, 5 API, 7 rate limited and 10 nothing configured, so a script branches on the number.

podcastindex-cli schema <command> prints the exact JSON Schema an MCP client receives for that tool.

MCP server, for your AI app

podcastindex-mcp is what Claude Code, Claude Desktop, Cursor and the rest launch. You never run it by hand:

claude mcp add podcastindex \
  -e PODCASTINDEX_API_KEY=your_key \
  -e PODCASTINDEX_API_SECRET=your_secret \
  -- npx -y @thenavidm/podcastindex-mcp-cli@latest

In Claude Desktop, the .mcpb extension installs on a double click. Section 4 has every other client.

A feed submission waits for your approval in the client, as section 9 explains.

Which one

They are the same program reading the same tool definitions, so anything one can do, the other can.

What each costs

Both surfaces are the same program with the same 36 tools. The difference is when the model pays for them. Measured in Claude Code:

Claude Code's tool search is on by default: it sends only the tool names and the server instructions, and loads a tool's full definition when the model reaches for it. An app that loads every tool up front pays the first line on every message, whether Podcast Index comes up or not. With the skill added, Claude Code also lists its one-line description, about 160 tokens.

To spend less, turn the server off when you are not using it, which in Claude Code is the /mcp panel. PODCASTINDEX_READ_ONLY=1 takes the 3 write tools off the list, leaving 33. Or install the CLI and add the server on the days it earns its place.

Measured on 2026-10-05 with Claude Code 2.1.286 on Claude Opus 5.5: one short prompt with and without the server connected, once with ENABLE_TOOL_SEARCH=false and once with the default, the difference read from the API's own usage figures. SKILL.md was measured the same way. Other apps and models count tokens a little differently.

Against 1.1.2, measured the same day: every tool loaded costs 11,843 tokens instead of 13,095, tool search the same (1,253 against 1,253), and SKILL.md 60 more, because it now says how approval works over MCP and lists every exit code. In Codex 0.159.3 on gpt-6.1-sol, the same task, "find the command that finds the moment a phrase was said in an episode's transcript, and the flags it requires", read a median of 83,167 input tokens on 2.0.0 against 83,990 on 1.1.2 over the CLI, five runs each: Codex now asks which instead of reading the full command list. Over MCP, Codex prints the tool list with a script and cuts the printout to about 10,000 tokens, so it read about 9,020 on both versions, out of a full listing of 28,195 tokens on 2.0.0 against 28,239, and a median of 48,367 input tokens against 48,444.

Features

Every tool is both a command and an MCP tool, with the same name. The command is the tool name with dashes.

All 36 are in section 6.

Contents

1. What you can ask it 💬