How dem108/mcp-foundry works
Project type and purpose
• This is an open-source, polyglot MCP (“Modular Chat Protocol”) server project.
• Purpose: provide interchangeable servers (in Python and TypeScript) that speak the MCP protocol, forwarding requests to Azure AI Foundry agent services and returning responses to any MCP-compatible client.
File structure
• Top-level: GitHub metadata, docs, CHANGELOG, LICENSE, README.
• docs/: Markdown describing each MCP server variant.
• src/python/:
• pyproject.toml (build config)
• azure_agent_mcp_server/ (package) with __main__.py entrypoint
• requirements.txt
• src/typescript/:
• package.json + package-lock.json (npm config)
• tsconfig.json (ESM + TS compiler settings)
• src/index.ts (server entrypoint)
• uv.lock (lockfile)
README for additional insights
• Confirms two server implementations (Azure AI Agent Service MCP Server).
• Describes polyglot structure and developer workflows for each language.
• States the servers adhere to the MCP protocol spec so any MCP client can connect.
Essential architectural elements
• Clearly separate the two implementations but group them under “MCP Server”
• Standard development flow for each (install → build → run).
• Round-trip data flow: Client → Server → Azure AI → Server → Client.