Firesidefm MCP Server

firesidefm-mcp-server

An MCP server for the Fireside.fm podcast hosting API. It lets an MCP host such as Claude Code or Claude Desktop read your podcasts, episodes and download statistics, create, edit, publish and unpublish episodes, and upload episode audio, cover art and transcripts.

This is an unofficial project. It is not affiliated with or endorsed by Fireside.

Requirements

• Node.js 22 or later
• A Fireside API key, created at https://app.fireside.fm/api_tokens

A Read Only key enables the read tools; a Read/Write key also enables the tools that change episodes. Use a Read Only key unless you want your assistant to create, edit or publish episodes. A key covers every podcast on your Fireside account. A key can be switched between Read Only and Read/Write under Account > API Keys; restart the server afterwards, since it chooses its tools when it starts.

Configuration

At startup the server checks the key with Fireside. The write tools are offered only when the key is Read/Write and FIRESIDE_READ_ONLY is not set. A rejected key stops the server with an explanation on stderr. If the key cannot be checked for any other reason, such as a network error or a Fireside outage, the server starts with only the read tools and says why on stderr; restart it to try again.

fireside_upload_file is offered only with a Read/Write key and FIRESIDE_UPLOAD_DIRS set. It sends only files inside those directories (symbolic links are followed before the check, which is repeated once the file is open; a hard link cannot be told apart from the file it points to, and another program rewriting those directories at the same moment could race the check), so set it to the folders where you keep finished episode files rather than your home directory.

Both hosts below store the key in plain text in their configuration file.

Claude Code

claude mcp add fireside \
  --env FIRESIDE_API_TOKEN=fire_… \
  --env FIRESIDE_PODCAST=your-podcast-slug \
  -- npx -y firesidefm-mcp-server

This adds the server for the current project only. Add --scope user to make it available in every project.

Claude Desktop

Edit Claude Desktop's configuration file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Add an entry under mcpServers (create the object if the file doesn't have one yet):

{
  "mcpServers": {
    "fireside": {
      "command": "npx",
      "args": ["-y", "firesidefm-mcp-server"],
      "env": {
        "FIRESIDE_API_TOKEN": "fire_…",
        "FIRESIDE_PODCAST": "your-podcast-slug"
      }
    }
  }
}

Then quit and reopen Claude Desktop.

Claude Desktop does not read your shell profile, so if Node was installed with a version manager such as nvm, it may not find npx. In that case set "command" to the full path printed by which npx.

Tools

Podcasts and episodes can be named by token or by slug. An episode's slug defaults to its episode number, so "episode 42" works unless the slug was customised in the dashboard (for example s02e04).

fireside_get_episode includes the episode's preview link, which plays the episode for anyone who has it, even before release.

Before uploading an MP3, the server reads its headers. A file that holds no MP3 audio (such as a renamed WAV) is refused. One that differs from the settings the Fireside dashboard suggests for MP3s (mono, 64–96 kb/s, 44.1 kHz) is stopped with a list of the differences; your assistant should ask you, and uploads it as it is only with accept_audio_specs: true. The audio limit depends on your Fireside plan, which the API does not report; Fireside checks the size before anything is sent, and the server then names the file's size as a likely cause.

Uploaded audio transfers to the episode in the background; audio_status on the episode shows processing, then ready or failed (with audio_error). Cover art and transcripts are attached during the call. Large audio files can take longer to upload than an MCP host allows for a tool call; in Claude Code, raise the limit with the MCP_TOOL_TIMEOUT environment variable (milliseconds).

Limitations

These come from the Fireside API itself:

• Chapters cannot be uploaded through the API. Publishing an episode without audio fails.
• Files must fit your plan's limits: cover art up to 1.91 MB, transcripts up to 195 KB, and audio within your plan's MP3 limit.
• Episodes cannot be deleted, and podcast settings cannot be changed.
• Publishing also posts to the podcast's connected social accounts when the episode is set to do so in the dashboard. The API cannot see or change that setting.
• Requests are limited to 120 per minute per key. The server waits out short rate-limit pauses and reports longer ones.

Development

npm install
npm run build       # compiles to dist/
npm run typecheck
npm test            # unit and in-process MCP tests; no network
npm run test:e2e    # builds, then runs dist/cli.js over stdio against a local stand-in API
npm run check       # all of the above
npm run dev         # run from source with tsx

scripts/ holds two checks that run against the real Fireside API with a key read from the macOS Keychain; see the comment at the top of each file. live-read-check.ts only reads. live-draft-check.ts creates and edits one draft episode, which you then delete in the dashboard; it refuses to send publish or unpublish requests.

License

MIT