Fathom MCP Server

Implementation Reference

• src/index.ts:132-142 (handler)

  The core execution logic (handler function) for the MCP 'get_transcript' tool. It fetches the transcript using the FathomClient and formats it to Markdown before returning as text content.

  async ({ recording_id }) => {
    console.error(`Fetching transcript for ${recording_id}...`);
    
    const response = await fathom.getTranscript(recording_id);
    const markdown = formatTranscriptToMarkdown(response.transcript);
    
    console.error('Transcript retrieved');
    return {
      content: [{ type: 'text', text: markdown }],
    };
  }

• src/index.ts:129-131 (schema)

  Input schema definition for the 'get_transcript' tool using Zod, requiring a numeric recording_id.

  {
    recording_id: z.number().describe('The recording ID of the meeting'),
  },

• src/index.ts:127-143 (registration)

  Registration of the 'get_transcript' MCP tool on the McpServer instance, including name, input schema, and handler reference.

  server.tool(
    'get_transcript',
    {
      recording_id: z.number().describe('The recording ID of the meeting'),
    },
    async ({ recording_id }) => {
      console.error(`Fetching transcript for ${recording_id}...`);
      
      const response = await fathom.getTranscript(recording_id);
      const markdown = formatTranscriptToMarkdown(response.transcript);
      
      console.error('Transcript retrieved');
      return {
        content: [{ type: 'text', text: markdown }],
      };
    }
  );

• src/fathom-client.ts:252-254 (helper)

  FathomClient helper method that makes the API request to retrieve the transcript for a specific recording ID.

  async getTranscript(recordingId: number): Promise<TranscriptResponse> {
    return this.request<TranscriptResponse>(`/recordings/${recordingId}/transcript`);
  }

• src/formatters.ts:81-103 (helper)

  Utility function to format the raw transcript array into readable Markdown with speaker labels and timestamps.

  export function formatTranscriptToMarkdown(transcript: TranscriptItem[]): string {
    if (!transcript || transcript.length === 0) {
      return '_No transcript available_';
    }
  
    const lines: string[] = [];
    let currentSpeaker = '';
  
    for (const item of transcript) {
      const speakerName = item.speaker.display_name || 'Unknown Speaker';
      
      // Add speaker header when speaker changes
      if (speakerName !== currentSpeaker) {
        currentSpeaker = speakerName;
        lines.push('');
        lines.push(`**${speakerName}** _[${formatTimestamp(item.timestamp)}]_`);
      }
      
      lines.push(`> ${item.text}`);
    }
  
    return lines.join('\n');
  }