MCP explained: the open protocol standardizing how AI agents connect to tools
robertonappe · x · 2026-10-26
A primer on the Model Context Protocol: an open standard introduced by Anthropic in November 2024 to connect AI agents and assistants with tools, data, and applications, replacing bespoke per-pair integrations. Technically it uses JSON-RPC 2.0 across hosts, clients, and servers, with servers exposing resources (data/context), prompts (work templates), and tools (executable functions)—modeled after the Language Server Protocol. In December 2025 Anthropic donated it to the Linux Foundation's Agentic AI Foundation (co-founded with Block and OpenAI); by then 10,000+ public MCP servers existed, adopted by ChatGPT, Gemini, Microsoft Copilot, Cursor, and VS Code.
More from coding & agent
- 5-layer agent memory setup: Mem0 cuts tokens 93%, Snowflake ontology lifts accuracy 20% — Arindam_1729 · 2026-10-26
- Skills written entirely by the model are useless, and stacking skills taxes agent context — HamelHusain · 2026-10-26
- Grok Bot hands-on: auto-logs hundreds of tasks to Linear and runs scheduled security checks — koltregaskes · 2026-10-26
- agent-saifu: an MCP server that gives AI agents their own isolated DeFi wallets — DecisiveCS · 2026-10-26
- GigaMail ships three MCP security patches against indirect prompt injection leaks — Soft-Lie-434 · 2026-10-26
- Hermes Agent Lands Web Desktop UI, Enabled with a Single Command Locally — intellectronica · 2026-10-26