MCP server catalogue
The Model Context Protocol is Anthropic's open standard for wiring tools, data, and prompts into agents. The spec is at modelcontextprotocol.io; servers exist for filesystems, git repos, databases, Slack, Notion, browser automation, calendars, and dozens of other surfaces. Cross-vendor support is broad — Claude, ChatGPT, VS Code, Cursor, Copilot Studio, and others all speak MCP. Below: what MCP is, the official Anthropic servers, where to find community servers (and trust gates), and how to build your own.
- MCP.1 What MCP is, plainly sourced
Model Context Protocol — what it is, why Anthropic kicked it off, what it actually does for an agent app, and why other vendors (OpenAI, Google, Microsoft) have adopted it. Plus the architecture in one diagram.
- MCP.2 Reference MCP servers sourced
The reference set of MCP servers at github.com/modelcontextprotocol/servers — what's actively maintained today (fetch · filesystem · git · memory · sequentialthinking · time · everything), what's been archived (github · postgres · sqlite · slack · gdrive · brave-search · puppeteer · redis), and the tool surface each one ships.
- MCP.3 Community MCP servers — where to find them, how to vet them sourced
Community-maintained MCP servers extend Claude / ChatGPT / Cursor / VS Code Copilot far beyond the official set. Where to discover them (smithery.ai, awesome-mcp lists, GitHub), trust signals to look for, and the questions to ask before installing one.
- MCP.4 Build your own MCP server sourced
Thirty-minute walkthrough — write an MCP server in Python (or TypeScript) that exposes one tool, wire it into Claude Code, and confirm it works. Plus where to extend from there (resources, prompts, multi-tool servers).