Pocket Casts Server

An MCP server that connects AI assistants to your Pocket Casts account. Browse your subscriptions, read episode details, and retrieve transcripts — with automatic transcription via AssemblyAI when no native transcript exists.

Built with Bun and FastMCP. Deployable to Railway or any Docker host.

Tools

Tool	Description
list-podcasts	List all subscribed podcasts (includes folder UUIDs)
new-episodes	Get recent episodes from your subscriptions
get-episode	Get details about a specific episode by UUID
get-transcript	Get an episode's transcript (falls back to AssemblyAI if none exists)
check-transcripts	Check transcript availability across all podcasts or a specific folder

Setup

Prerequisites

• Bun v1.3+
• A Pocket Casts account (Plus required for API access)

Install and authenticate

bun install
bun run login

The login command will prompt for your Pocket Casts email and password, then save tokens to auth.json.

Configure environment

Create a .env file:

MCP_API_KEY=your-secret-key-here
ASSEMBLYAI_API_KEY=your-assemblyai-key    # optional, enables transcription fallback

Run

bun run start

The server starts on port 3001 (override with PORT env var) using HTTP stream transport.

Connecting to an MCP client

Point your MCP client at:

http://localhost:3001/mcp

Authenticate with either:

• Authorization: Bearer <MCP_API_KEY> header
• ?api_key=<MCP_API_KEY> query parameter

Deploy to Railway

The included Dockerfile is Railway-ready. Set these environment variables in your Railway project:

Variable	Required	Description
MCP_API_KEY	Yes	Bearer token for MCP client auth
POCKETCASTS_ACCESS_TOKEN	Yes	From auth.json after login
POCKETCASTS_REFRESH_TOKEN	Yes	From auth.json after login
POCKETCASTS_EXPIRES_AT	Yes	From auth.json after login
ASSEMBLYAI_API_KEY	No	Enables transcription for episodes without native transcripts
PORT	No	Railway sets this automatically

How transcripts work

1. Checks for Pocket Casts-generated transcripts first
2. Falls back to RSS-sourced transcripts if available
3. If neither exists and ASSEMBLYAI_API_KEY is set, submits the audio URL directly to AssemblyAI for transcription (no file download required)