loading…
Search for a command to run...
loading…
▸ TL;DR
Function calling is a vendor-specific LLM API (OpenAI, Anthropic). MCP is an open protocol — same MCP server works with any client (Claude Desktop, Cursor, VS Code). MCP avoids vendor lock-in.
Read MCP explainerFunction calling is a feature of specific LLM APIs: you describe tools in OpenAI's JSON schema, call the API, the LLM picks a tool to invoke. It only works with that vendor and runs inside your application code. MCP is an inter-process protocol: a separate MCP server exposes tools, and any MCP-aware client (regardless of which LLM it uses internally) can call them. This means: one MCP server (say, GitHub MCP) works in Claude Desktop, Cursor, VS Code, and other clients. With function calling, you'd have to re-integrate for each app. MCP also handles streaming, cancellation, capability negotiation — features you'd have to build yourself with function calling.
Partially. MCP solves "tool plumbing" — connecting AI to tools with one protocol. LangChain is a broader agent framework with chains, memory, RAG. Use MCP if you want simple, language-agnostic tool integration. Use LangChain if you need a full agent runtime in Python/JS.
Yes, when you use MCPs from a vetted catalog. Unyly auto-scans every MCP for leaked secrets and malicious dependencies. Avoid installing random GitHub gists without verification.