Call.md Documentation

Call.md supports the Model Context Protocol (MCP) to allow an AI agent to trigger tools automatically during meetings. To connect your own MCP servers:

1. Navigate to Settings → MCP Servers within the application.
2. Click Add Server.
3. Select the appropriate transport type:
   • stdio: For local servers.
   • http: For remote servers.
4. Configure the server details and click Connect.

Once connected, the MCP agent will automatically detect information needs from the conversation and trigger relevant tools. Results will be displayed in the MCP Results panel during the meeting.

If you need assistance or want to engage with the community, use the following resources:

• Documentation: Detailed guides and API references are available at docs.videodb.io.
• Issues: Report bugs or request features on GitHub Issues.
• Discord: Join the community discussion on Discord.
• API Key: Manage your credentials and access your account via the VideoDB Console.

To install Call.md on macOS (supporting both Apple Silicon and Intel), run the following command in your terminal:

curl -fsSL https://artifacts.videodb.io/call.md/install | bash

Post-installation steps:

1. Launch Call.md from your Applications folder or via Spotlight.
2. Grant the required system permissions: Microphone and Screen Recording.
3. Register the application using your VideoDB API key (available at console.videodb.io).

Developers can run Call.md locally by following these steps:

1. Clone the repository:

   git clone https://github.com/video-db/call.md.git
   cd call-md

2. Install dependencies:

   npm install

3. Rebuild native modules for Electron:

   npm run rebuild

4. Start development mode:

   npm run dev

5. Register: Enter your VideoDB API key when the application opens.

Prerequisites:

• Node.js 18+
• npm 10+
• macOS 12+ (Monterey or later)

The ToolAggregatorService aggregates tools from all connected Model Context Protocol (MCP) servers. To prevent naming conflicts between different servers, it uses a namespacing pattern where tools are identified by the format serverId:toolName.

This service provides several ways to interact with tools:

• Namespaced Access: Using serverId:toolName to ensure you are calling the correct tool from a specific server.
• Simple Name Access: Searching by the tool's base name (returns the first match).
• Pattern/Keyword Search: Finding tools via regex patterns, keyword searches, or predefined intents (like crm, calendar, or docs).
• Execution: Running tools via executeTool using their namespaced name.

// Example of the namespacing concept
// If server 'google-cal' has a tool 'create_event'
// The namespacedName is 'google-cal:create_event'

The Markdown export service organizes data into a time-based folder hierarchy within ~/.call_md/. This allows for easy chronological browsing and prevents filename collisions.

Hierarchy:

1. Root: ~/.call_md/ contains the index.md file.
2. Meetings Root: ~/.call_md/meetings/
3. Date Folders: Nested folders for Year/Month/Day.
4. Meeting Folder: A sanitized version of the meetingName.

Example Path: ~/.call_md/meetings/2024/03/24/project-sync/

The useCopilot hook automatically manages IPC event listeners via window.electronAPI.copilotOn. When these events are received, the hook updates the internal React state. You do not need to manually subscribe to these events if you are using the hook.

Handled Events

• onTranscript: Receives new transcript segments and adds them to the state.
• onMetrics: Receives { metrics, health } updates to refresh real-time meeting data.
• onNudge: Receives a { nudge } object when a new suggestion is available.
• onCallEnded: Receives { summary, metrics, duration } to populate the post-meeting intelligence.
• onError: Logs errors received from the copilot backend.

Recording not starting

• Verify that Microphone and Screen Recording permissions are granted in macOS System Settings > Privacy & Security.
• Ensure your VideoDB API key is valid.

Transcription not appearing

• Confirm that both mic and system audio are enabled in the application settings.
• Allow 5-10 seconds for the first transcripts to appear.
• Check your internet connectivity (VideoDB features require an active connection).

Development issues

• Run npm run rebuild to fix native module issues.
• Ensure you are using Node.js 18+.
• Review application logs located at: ~/Library/Application Support/call-md/logs/.

The project uses Drizzle ORM with a SQLite dialect. To configure the database connection and schema management, use defineConfig from drizzle-kit.

Key configuration properties:

• schema: The path to your TypeScript schema definition file.
• out: The directory where Drizzle will output migrations.
• dialect: Set to 'sqlite' for this project.
• dbCredentials: An object containing connection details. For SQLite, use the url key to specify the path to the database file.

import { defineConfig } from 'drizzle-kit';

export default defineConfig({
  schema: './src/main/db/schema.ts',
  out: './drizzle',
  dialect: 'sqlite',
  dbCredentials: {
    url: './data/call-md.db',
  },
});

Call.md uses a local-first approach with a SQLite database. Data is stored at:

• Database: ~/Library/Application Support/call-md/data/call-md.db
• Logs: ~/Library/Application Support/call-md/logs/app-YYYY-MM-DD.log