Fathom MCP Server

A read-only Model Context Protocol (MCP) server for accessing Fathom AI API endpoints (meetings, recordings, transcripts, summaries, teams, team members) via GET operations. Built with FastMCP.

This implementation provides streamlined access to Fathom meeting data while minimizing API consumption. It is optimized for efficiency and simplicity, using the toon output format for less token usage and better LLM processing.

Features

• List Meetings: Retrieve meetings with optional filtering and inclusion of summaries

• Get Meeting Details: Retrieve comprehensive meeting data including AI-generated summaries and transcripts

• List Teams: Retrieve all teams

• List Team Members: Retrieve team members with optional team filtering

Requirements

• Python 3.10+

• Fathom API key

• FastMCP 2.0+

Installation

1. Clone or download this repository

2. Install dependencies:

or

Configuration

The server uses environment variables for configuration:

• FATHOM_API_KEY: Your Fathom API key (required)

• FATHOM_TIMEOUT: Request timeout in seconds (default: 30)

• OUTPUT_FORMAT: Output format for tool responses ("toon" or "json", default: "toon")

Usage

Direct Python Execution (Recommended)

Using UV

Available Tools

list_meetings

Retrieve meetings with optional filtering and pagination.

Properties:

• calendar_invitees (list[str], optional): Filter by invitee emails

• calendar_invitees_domains (list[str], optional): Filter by domains

• created_after (str, optional): ISO timestamp filter

• created_before (str, optional): ISO timestamp filter

• cursor (str, optional): Pagination cursor

• include_action_items (bool, optional): Include action items

• include_crm_matches (bool, optional): Include CRM matches

• per_page (int, optional): Number of results per page (default: 20, configurable via DEFAULT_PER_PAGE env var)

• recorded_by (list[str], optional): Filter by recorder emails

• teams (list[str], optional): Filter by team names

get_meeting_details

Retrieve comprehensive meeting details including summary and metadata (without transcript).

Properties:

• recording_id (int): The recording identifier

Returns: A unified meeting object containing:

• recording_id: Unique identifier for the recording

• title: Meeting title

• meeting_url: URL to the meeting recording

• share_url: Shareable URL for the meeting

• created_at: When the meeting was created

• scheduled_start_time: Original scheduled start time

• scheduled_end_time: Original scheduled end time

• recording_start_time: When recording actually started

• recording_end_time: When recording actually ended

• transcript_language: Language of the transcript

• participants: List of meeting participants with names, emails, and external/internal status

• recorded_by: Information about who recorded the meeting (name, email, team)

• teams: Teams associated with the meeting

• topics: AI-detected topics discussed

• sentiment: Overall sentiment analysis

• crm_matches: CRM contact matches

• summary: AI-generated meeting summary (converted to plain text from markdown)

get_meeting_transcript

Retrieve meeting transcript with essential metadata (id, title, participants, dates).

Properties:

• recording_id (int): The recording identifier

Returns: A transcript object containing:

• recording_id: Unique identifier for the recording

• title: Meeting title

• participants: List of meeting participants

• created_at: When the meeting was created

• scheduled_start_time: Original scheduled start time

• scheduled_end_time: Original scheduled end time

• transcript: Full meeting transcript with timestamps

list_teams

Retrieve teams with optional pagination.

Properties:

• cursor (str, optional): Pagination cursor

• per_page (int, optional): Number of results per page (default: 20, configurable via DEFAULT_PER_PAGE env var)

list_team_members

Retrieve team members with optional filtering and pagination.

Properties:

• cursor (str, optional): Pagination cursor

• per_page (int, optional): Number of results per page (default: 20, configurable via DEFAULT_PER_PAGE env var)

• team (str, optional): Filter by team name

Output Format

The server supports two output formats configured via the OUTPUT_FORMAT environment variable:

• TOON (default): Token-Optimized Object Notation - optimized for LLM processing with reduced token usage

• JSON: Standard JSON format with indentation for human readability

All output is filtered to remove empty, null, or redundant information for improved efficiency.

Error Handling

The server provides comprehensive error handling:

• 401 Unauthorized: Invalid API key

• 404 Not Found: Resource not found

• 429 Rate Limited: Too many requests

• 500 Server Error: Fathom API issues

All errors are logged via MCP context with appropriate severity levels.

Security

• API keys are loaded from environment variables

• No sensitive data is logged

• HTTPS is used for all API requests

• Error messages don't expose internal details

License

MIT License.