How echelon-ai-labs/zoom-mcp works
Main Components
• CLI Entry Point
• scripts/setup.sh, scripts/check_env.py, scripts/test_zoom_connection.py
• src/zoom_mcp/cli.py
• MCP Server Core
• src/zoom_mcp/server.py
• Authentication Module
• src/zoom_mcp/auth/zoom_auth.py
• Resource Handlers
• src/zoom_mcp/resources/recordings.py
• (future: other resources/*.py)
• Utility Tools
• src/zoom_mcp/tools/recordings.py
• Configuration
• .env (populated from .env.example)
• pyproject.toml (dependencies & entry points)
• External Services
• Zoom OAuth2 endpoint
• Zoom REST API
Relationships & Data Flows
• User invokes CLI or directly starts the MCP server.
• CLI → MCP Server: parses arguments, passes control to server.py.
• MCP Server receives HTTP requests (from Claude or other clients).
• For each request:
a. Server calls Auth Module to acquire/refresh OAuth token.
b. Auth Module ↔ Zoom OAuth2 endpoint (Basic Auth → access token).
c. Server calls Resource Handlers, passing token.
d. Resource Handler ↔ Zoom REST API (e.g., GET /meetings, /recordings).
e. Utility Tools process or transform API payloads as needed.
• Server returns JSON responses to client.
Architectural Patterns & Principles
• Layered Design:
• Presentation Layer (CLI & HTTP API)
• Service Layer (server.py orchestrator)
• Integration Layer (Auth + Resources + Tools)
• Single Responsibility: each module does one thing (auth, resources, tools).
• Configuration‐driven: credentials in .env, tool behavior via CLI flags.
• Extensibility: new resources or tools can be added in resources/ and tools/.
Capturing Essential Details
• .env loading by Auth Module.
• Token caching/renewal loop.
• How tests (tests/test_zoom_auth.py) simulate or mock Auth.
• Avoid unnecessary fragmentation—group closely related modules in one container.