Glossary · Definition

MCP (Model Context Protocol)

MCP (Model Context Protocol) is an open standard for connecting AI assistants to external tools and data sources. Think USB-C for AI integrations: write a server once, it works in Claude, ChatGPT, Cursor, Zed, Goose, etc.

Updated May 2026 · 4 min read

100% in-browserNo downloadsNo sign-upMalware-freeHow we keep this safe →

Definition

What it means

Anthropic launched MCP in late 2024. Servers expose three primitives: resources (read-only data), tools (callable functions), prompts (templates). Clients connect over JSON-RPC — stdio for local, HTTP/SSE for remote. Authentication is the server's responsibility. By 2026 there are hundreds of MCP servers — official ones for filesystem, GitHub, Postgres, Slack, plus vendor-published ones from Notion, Linear, Stripe, etc.

Why it matters

Before MCP, every AI integration was custom — your Slack-Claude bot couldn't be reused for Cursor. MCP unified this. If you're building with AI in 2026, learning MCP is essential — it's how you wire up 'agent + my company data + my external services' without writing N integrations.

Related free tools

Free toolMCP Server PickerPick a workflow (coding, research, data, support, scraping) and get the recommended MCP servers plus a copy-paste Claude Desktop config.Open tool →

Frequently asked questions

Where do I find MCP servers?

github.com/modelcontextprotocol/servers (official), or browse community MCP marketplaces. Use our MCP server picker for workflow-based recommendations.

Can I write my own?

Yes — the protocol is well-documented at modelcontextprotocol.io. Python and TypeScript SDKs available.

Related terms

DefinitionAI agentAn AI agent is an LLM running in a loop: think → call a tool → observe the result → think again. The loop continues until the task is done or a stopping condition is hit.