fathom-mcp-server

# Fathom AI MCP Server - Complete Documentation

> A remote MCP server that connects AI assistants like Claude to Fathom AI meetings, transcripts, and AI summaries.

This is the full documentation for LLMs and AI systems. For a shorter version, see /llms.txt

## Overview

Fathom AI MCP Server is an open source Model Context Protocol (MCP) server that enables MCP hosts (like Claude Desktop) to access Fathom AI meeting data. It acts as a secure bridge between AI assistants and Fathom AI's meeting recording and transcription service.

Think of MCP as "USB-C for AI applications" — this server plugs Fathom's capabilities into any MCP-compatible AI assistant.

## Server Information

- Type: Remote MCP Server
- Transports: Streamable HTTP, SSE (legacy)
- Authentication: OAuth 2.0
- Streamable HTTP URL: https://www.fathom-mcp-server.com/mcp
- SSE URL: https://www.fathom-mcp-server.com/sse
- API Docs: https://www.fathom-mcp-server.com/docs
- Source: https://github.com/agencyenterprise/fathom-mcp-server
- License: MIT

## MCP Capabilities

- Tools: Yes (6 tools)
- Resources: No
- Prompts: No

## Compatible MCP Hosts

- Claude Desktop
- Claude.ai (via Connectors)
- Perplexity Desktop (via SSE connector)
- Any MCP-compatible AI assistant or agent

---

## MCP Tools - Complete Reference

### 1. list_meetings

List meetings with optional filters including date range, team, and recorder.

**Parameters:**

| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| calendar_invitees_domains | string[] | No | Filter by attendee email domains |
| calendar_invitees_domains_type | enum | No | "all", "only_internal", or "one_or_more_external" |
| created_after | ISO datetime | No | Only meetings created after this date |
| created_before | ISO datetime | No | Only meetings created before this date |
| cursor | string | No | Pagination: pass the next_cursor from the previous response to get the next page |
| include_action_items | boolean | No | Include action items in response |
| include_crm_matches | boolean | No | Include CRM matches in response |
| recorded_by | string[] | No | Filter by recorder email addresses |
| teams | string[] | No | Filter by team names |

**Response Schema:**

```json
{
  "limit": number | null,
  "next_cursor": string | null,
  "items": [
    {
      "title": string,
      "meeting_title": string | null,
      "recording_id": number,
      "url": string,
      "share_url": string,
      "created_at": string,
      "scheduled_start_time": string,
      "scheduled_end_time": string,
      "recording_start_time": string,
      "recording_end_time": string,
      "calendar_invitees_domains_type": "only_internal" | "one_or_more_external",
      "transcript_language": string,
      "calendar_invitees": [
        {
          "name": string | null,
          "email": string | null,
          "email_domain": string | null,
          "is_external": boolean,
          "matched_speaker_display_name": string | null
        }
      ],
      "recorded_by": {
        "name": string,
        "email": string,
        "email_domain": string,
        "team": string | null
      },
      "transcript": [...] | null,
      "default_summary": {...} | null,
      "action_items": [...] | null,
      "crm_matches": {...} | null
    }
  ]
}
```

**Pagination:** When `next_cursor` is non-null, call list_meetings again with `cursor` set to that value to fetch more meetings.

**Example Usage:**

```
User: "Show me my meetings from last week"
Tool call: list_meetings with created_after="2026-01-27T00:00:00Z", created_before="2026-02-03T23:59:59Z"
```

---

### 2. search_meetings

Search meetings by title. This is an MCP-native tool that performs client-side filtering since Fathom's API doesn't provide a search endpoint.

**Parameters:**

| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| query | string | Yes | Search term to match against meeting title |
| calendar_invitees_domains | string[] | No | Filter by attendee email domains |
| calendar_invitees_domains_type | enum | No | "all", "only_internal", or "one_or_more_external" |
| created_after | ISO datetime | No | Only meetings created after this date |
| created_before | ISO datetime | No | Only meetings created before this date |
| cursor | string | No | Pagination: pass next_cursor from the previous response to get the next page |
| include_action_items | boolean | No | Include action items |
| include_crm_matches | boolean | No | Include CRM matches |
| recorded_by | string[] | No | Filter by recorder emails |
| teams | string[] | No | Filter by team names |

**Response Schema:**

Same as list_meetings, but filtered to only include meetings where the title or meeting_title contains the query string (case-insensitive). When `next_cursor` is non-null, call search_meetings again with the same query and filters and `cursor` set to that value to search more pages.

**Performance Note:**

For users with many meetings, this tool fetches all meetings matching the filters then filters client-side. For better performance, use list_meetings with date filters instead.

**Example Usage:**

```
User: "Find meetings about the product roadmap"
Tool call: search_meetings with query="product roadmap"
```

---

### 3. get_transcript

Get the full transcript for a specific recording.

**Parameters:**

| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| recording_id | string | Yes | The recording ID (from list_meetings response) |

**Response Schema:**

```json
{
  "transcript": [
    {
      "speaker": {
        "display_name": string,
        "matched_calendar_invitee_email": string | null
      },
      "text": string,
      "timestamp": string
    }
  ]
}
```

**Example Usage:**

```
User: "Get the transcript from my standup yesterday"
Steps:
1. Call list_meetings with date filter for yesterday
2. Find the standup meeting in results
3. Call get_transcript with recording_id from that meeting
```

---

### 4. get_summary

Get the AI-generated summary for a specific recording.

**Parameters:**

| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| recording_id | string | Yes | The recording ID (from list_meetings response) |

**Response Schema:**

```json
{
  "summary": {
    "template_name": string | null,
    "markdown_formatted": string | null
  }
}
```

The `markdown_formatted` field contains the AI-generated summary in markdown format, typically including:
- Key discussion points
- Decisions made
- Action items
- Next steps

**Example Usage:**

```
User: "Summarize my meeting with the design team"
Steps:
1. Call search_meetings with query="design team" or list_meetings with date filters
2. Find the relevant meeting
3. Call get_summary with recording_id from that meeting
```

---

### 5. list_teams

List all teams the authenticated user has access to.

**Parameters:**

| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| cursor | string | No | Pagination cursor for next page |

**Response Schema:**

```json
{
  "items": [
    {
      "name": string,
      "created_at": string
    }
  ],
  "limit": number,
  "next_cursor": string | null
}
```

**Example Usage:**

```
User: "What teams do I have access to?"
Tool call: list_teams
```

---

### 6. list_team_members

List all members of a specific team.

**Parameters:**

| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| team | string | No | Team name to list members for |
| cursor | string | No | Pagination cursor for next page |

**Response Schema:**

```json
{
  "items": [
    {
      "name": string,
      "email": string,
      "created_at": string
    }
  ],
  "limit": number,
  "next_cursor": string | null
}
```

**Example Usage:**

```
User: "Who is on the engineering team?"
Tool call: list_team_members with team="Engineering"
```

---

## Common Workflows

### Workflow 1: Review Yesterday's Meetings

1. Call `list_meetings` with `created_after` set to yesterday's date
2. Present the list of meetings to the user
3. If user wants details, call `get_transcript` or `get_summary` for specific recordings

### Workflow 2: Find Specific Meeting Content

1. Call `search_meetings` with the user's search term
2. Present matching meetings
3. Call `get_transcript` to get full conversation text
4. Analyze or summarize the transcript content

### Workflow 3: Action Item Follow-up

1. Call `list_meetings` with `include_action_items=true`
2. Extract and present action items from recent meetings
3. Help user track completion status

### Workflow 4: Team Meeting Overview

1. Call `list_teams` to get team list
2. Call `list_meetings` with `teams` filter for specific team
3. Summarize team's meeting activity

---

## Error Handling

All tools return errors in this format when something goes wrong:

```json
{
  "content": [{ "type": "text", "text": "Error message here" }],
  "isError": true
}
```

Common errors:
- "Invalid parameters" - Check parameter types and required fields
- "Unauthorized" - OAuth token expired, user needs to re-authenticate
- "Not found" - Recording ID doesn't exist or user doesn't have access

---

## Security & Privacy

### Data Flow

1. User authenticates via OAuth with Fathom
2. MCP server stores only encrypted OAuth tokens
3. Meeting data flows: Fathom API → MCP Server → Claude (pass-through, never stored)

### What's Stored

| Data | Stored | Encrypted |
|------|--------|-----------|
| OAuth tokens | Yes | Yes (AES-256-GCM) |
| Meetings | No | N/A |
| Transcripts | No | N/A |
| Summaries | No | N/A |

### Privacy Policy

Full policy: https://www.fathom-mcp-server.com/privacy

---

## How to Connect

### For Individual Users

1. Open Claude Desktop
2. Go to Settings > Connectors > Add Custom Connector
3. Paste: https://www.fathom-mcp-server.com/mcp
4. Click connect and authenticate with Fathom

For SSE clients (e.g. Perplexity Desktop):

1. Use the SSE URL: https://www.fathom-mcp-server.com/sse
2. Refer to your client's documentation for how to add a remote SSE connector
3. Authenticate with your Fathom account via OAuth

### For Organizations

Admins must add the connector via organization admin settings (not personal settings) for it to be available to all organization members.

---

## Example Conversation

**User:** "What meetings did I have this week?"

**Assistant:** Let me check your meetings from this week.

*[Calls list_meetings with created_after="2026-02-03T00:00:00Z"]*

You had 5 meetings this week:
1. Monday 10am - Weekly Standup (45 min)
2. Monday 2pm - Product Review (1 hour)
3. Tuesday 11am - Design Sync (30 min)
4. Wednesday 3pm - Client Call - Acme Corp (45 min)
5. Thursday 9am - Sprint Planning (1.5 hours)

Would you like me to get the transcript or summary from any of these?

**User:** "Get me the summary from the client call"

**Assistant:** *[Calls get_summary with recording_id from the Acme Corp meeting]*

Here's the summary from your client call with Acme Corp:

**Key Discussion Points:**
- Reviewed Q1 deliverables and timeline
- Discussed new feature requests for mobile app
- Addressed concerns about API performance

**Decisions Made:**
- Agreed to extend deadline by 2 weeks
- Will prioritize mobile notifications feature

**Action Items:**
- Send updated timeline by Friday (you)
- Provide API usage metrics (client)
- Schedule follow-up for next Tuesday

---

## Related Resources

- Model Context Protocol: https://modelcontextprotocol.io
- MCP Specification: https://spec.modelcontextprotocol.io
- MCP Registry: https://modelcontextprotocol.io/registry
- Fathom: https://fathom.video
- Fathom API Docs: https://developers.fathom.ai
- Fathom LLMs.txt: https://developers.fathom.ai/llms.txt

---

## Support & Contributing

- GitHub Issues: https://github.com/agencyenterprise/fathom-mcp-server/issues
- Contributing Guide: https://github.com/agencyenterprise/fathom-mcp-server/blob/main/CONTRIBUTING.md

## License

MIT License - https://github.com/agencyenterprise/fathom-mcp-server/blob/main/LICENSE