Skip to content

agencyenterprise/fathom-mcp-server

Repository files navigation

Fathom AI MCP Server

Connect Claude to your Fathom meetings, transcripts, and AI summaries.

License: MIT Node.js MCP MCP Registry

Get Started | Available Tools | API Docs | Self-Hosting | Contributing

Security


Get Started

This server is published to the MCP Registry as io.github.agencyenterprise/fathom-mcp-server.

Connect in under 60 seconds:

https://www.fathom-mcp-server.com/mcp
  1. Open Claude Desktop
  2. Go to Settings > Connectors > Add Custom Connector
  3. Paste the URL above
  4. Authenticate with Fathom

That's it. Ask Claude about your meetings.

Organizations: Admins must add the connector via organization admin settings, not personal settings.

Available Tools

Tool Description Docs
list_meetings List meetings with filters (date, team, recorder, etc.) Fathom API
search_meetings Search meetings by title with optional filters MCP Custom
get_transcript Get full transcript for a recording Fathom API
get_summary Get AI-generated summary for a recording Fathom API
list_teams List all accessible teams Fathom API
list_team_members List members of a team Fathom API

Custom MCP Tools

search_meetings

Search Fathom meetings by title or meeting_title. This is an MCP-native tool that performs client-side filtering since Fathom's API doesn't provide a search endpoint. For users with many meetings, use list_meetings with date filters for better performance.

See the Fathom MCP Server documentation for full request and response parameters.

Example Usage in Claude

"Show me my meetings from last week"

"Get the transcript from my standup yesterday"

"Summarize my meeting with the design team"

Security

Your Fathom data is never stored or exposed by this server.

Data Stored? Exposed? Details
Meetings, transcripts, summaries No No Fetched from Fathom and passed directly to Claude
Team and member info No No Fetched from Fathom and passed directly to Claude
OAuth tokens Yes No Encrypted at rest (AES-256-GCM), never logged or exposed
  • Pass-through architecture: This server acts as a secure proxy, your Fathom data flows directly from Fathom to Claude without being stored, cached, or logged
  • Encryption at rest: The only stored data (OAuth tokens) is encrypted using AES-256-GCM before being written to the database
  • HTTPS only: All communication between Claude, this server, and Fathom is encrypted in transit

See our full Privacy Policy for details on data collection, usage, and your rights.

Permissions

This MCP server defines a custom scope called fathom:read for tokens it issues to Claude. This is not a Fathom API scope - it's specific to this MCP server to describe read-only access to your Fathom data.

The Fathom API itself only provides read access via its public_api scope. Write operations (creating/editing meetings, transcripts, etc.) are not available in the Fathom API.

Limitations

  • search_meetings performs client-side filtering since Fathom's API doesn't provide a search endpoint. For users with many meetings, use list_meetings with date filters instead.
  • You can always ask the LLM what query params are avaialable.

Self-Hosting

Fathom OAuth apps require HTTPS redirect URIs, so local development with http://localhost isn't possible. Deploy to a hosting provider to test.

1. Deploy to a Hosting Provider

Railway (recommended), Render, or any platform that provides:

  • Node.js 18+ runtime
  • PostgreSQL database
  • HTTPS URL

Railway setup:

  1. Fork/clone this repo
  2. Create a new Railway project (you can deploy directly from your forked Github repo)
  3. Add a PostgreSQL database service in project
  4. Connect Database url to deployed repo in project and setup other envs

2. Create a Fathom OAuth App

  1. Go to Fathom Developer Portal
  2. Click "Register your app" (requires Fathom admin access)
  3. Set the redirect URI to https://your-app-url.railway.app/oauth/fathom/callback
  4. Note your Client ID and Client Secret

3. Configure Environment Variables (locally and in Railway)

Set these in your hosting provider's dashboard (as well as your local .env file to test build and start commands locally before pushing changes)

Variable Description
DATABASE_URL PostgreSQL connection string (auto-set by Railway - use public db url)
BASE_URL Your app's public URL (e.g., https://your-app.railway.app)
TOKEN_ENCRYPTION_KEY 32-byte hex key (generate with openssl rand -hex 32)
FATHOM_CLIENT_ID From step 2
FATHOM_CLIENT_SECRET From step 2

4. Initialize Database

Run migrations after first deploy:

npm run db:migrate

Or via Railway CLI:

railway run npm run db:migrate

5. Connect Claude

Add your deployed URL as a custom connector in Claude Desktop:

https://your-app.railway.app/mcp

Development

npm run dev          # Start dev server with hot reload only for testing
npm run build        # Build for production
npm run start        # Run production build
npm run lint         # Check for linting errors
npm run lint:fix     # Fix linting errors
npm run typecheck    # Run TypeScript type checking
npm run ci           # Run all CI checks (lint, typecheck, build)
npm run format       # Format code with Prettier
npm run db:studio    # Open Drizzle Studio for database inspection
npm run db:generate  # Generate migrations from schema changes
npm run db:migrate   # Run pending migrations
npm run db:push      # Push schema directly (dev only)

Tip: Run npm run ci before pushing to ensure your changes pass GitHub Actions.

Beta Testing

For pre-release features, use the staging URL:

https://fathom-mcp-staging.up.railway.app/mcp

Fathom AI Deep dive

https://developers.fathom.ai/llms.txt

Contributing

See CONTRIBUTING.md for development setup and guidelines.

Releasing

See RELEASING.md for version and release instructions.

Future Development Plans

  • Transcript vectorization — Enable vectorization of large transcripts so LLMs can parse and understand them more efficiently. Would be implemented as a stateless worker to ensure no user data is persisted.
  • Action item aggregation — Aggregate action items across meetings with filters. "Show all my incomplete action items from this week."
  • Meeting analytics — Calculate stats like total meeting time, meeting frequency, and top attendees.
  • Speaker time analysis — Analyze transcripts to show who spoke most in a meeting.
  • Meeting comparison — Compare two meeting summaries to highlight what changed over time.
  • Fathom API changelog monitoring — Automated detection of Fathom API changes via GitHub Action that periodically checks their API reference and creates an issue if changes are detected.

Contributions toward these goals are welcome. See CONTRIBUTING.md.

License

MIT License - see LICENSE for details.