The open protocol connecting LLMs with tools and data
Introduction
MCP (Model Context Protocol) is the open standard from Anthropic for connecting LLMs to data, tools and services — the 'USB-C of AI apps.' Adopted by OpenAI, Google and other major vendors, with a booming server ecosystem, it is becoming the de facto interconnect for agents.
Key Features
- Unified tools/resources/prompts spec
- Broad adoption across major vendors
- Booming server ecosystem
- Local and remote transports
- Official multi-language SDKs
Pros
- First-mover ecosystem at standard status
- Ends fragmented function-calling schemes
Cons
- Spec still evolving quickly
Pricing
The protocol and SDKs are fully open and free.（See official site for current pricing）
FAQ
How to write an MCP server?
Official Python/TS SDKs expose tools in a few dozen lines.
vs Function Calling?
FC is vendor-private; MCP is an open protocol across models and apps.
Where to find servers?
The official registry and community Awesome lists.
Related Sites
Collected on lucifer.city · Reviewed by the founder