MCP
MCP is an open protocol that lets LLM agents connect to external tools, data sources, and services through a standardized server interface.
The Model Context Protocol (MCP) is an open protocol introduced by Anthropic that standardizes how LLM-powered agents connect to external tools, data sources, and services. Instead of each AI tool inventing its own plugin API, MCP defines a shared wire format — so one MCP server (for, say, a Postgres database or a Jira instance) works with every MCP-aware client.
Why it matters
Before MCP, every agentic coding CLI had to reinvent integrations. Claude Code had its plugin format, Codex had another, an IDE plugin had a third. MCP collapses this to one interface: an MCP server exposes tools and resources, any client that speaks MCP can call them. This is why Claude Code, Codex CLI, Qwen Code, and Kimi CLI can all plug into the same Postgres, GitHub, or browser-automation servers.
For developers building AI workflows, MCP means you write one integration and it works with every tool. For users, MCP means you configure a server once (in your client's config) and it's available in every session.
How it works
MCP is a JSON-RPC protocol usually run over stdio or HTTP/SSE. Three primitives:
- Tools — callable functions the model can invoke via tool use
- Resources — readable data the agent can request (files, rows, pages)
- Prompts — parameterized prompt templates the client can surface as slash commands
A typical session: the client launches one or more MCP servers as subprocesses. Each server advertises its capabilities during the handshake. When the LLM wants to call a tool, the client forwards the call to the right server and streams the result back into the context window.
Official SDKs exist in TypeScript, Python, Rust, and others. Anthropic and community maintainers publish servers for GitHub, Slack, Postgres, Puppeteer, filesystem access, and many more.
How it's used
Common MCP patterns:
- Give Claude Code read access to your production Postgres via the
postgresMCP server - Let Codex CLI drive a browser via
puppeteer/playwrightMCP servers - Share a single "company knowledge" MCP server across your whole team
- Build custom MCP servers for internal systems
SpaceSpider doesn't intercept MCP — it's configured per-CLI via each tool's own config file (~/.claude/, ~/.codex/, etc.).
Related terms
- Tool use — the mechanism MCP enables
- Claude Code — MCP's reference client
- Subagent — often paired with custom MCP servers
- Hook — a different extensibility mechanism
- Autonomous agent — the context MCP lives in
FAQ
Is MCP an Anthropic-only thing?
No. Anthropic released the spec as an open standard in late 2024 and many non-Anthropic clients now support it, including Codex CLI, Qwen Code, Zed, Cursor, and others.
Do I need MCP to use Claude Code?
No. Claude Code works out of the box with built-in tools. MCP is for connecting to external systems beyond your local filesystem and shell.
Related terms
- Agentic codingAgentic coding is software development where an LLM-powered agent plans, edits, runs, and verifies code on its own using tools, not just autocomplete.
- AI pair programmingAI pair programming is a collaboration style where an LLM assistant sits alongside you, suggesting code and reviewing changes in real time as you work.
- ANSI escape codesANSI escape codes are control sequences that terminals interpret for colors, cursor movement, and screen clearing — the language of every modern CLI UI.
- Autonomous agentAn autonomous agent is an AI program that perceives, decides, and acts on its own toward a goal — the architecture behind modern coding CLIs.
- CheckpointA checkpoint is a saved snapshot of file state that lets you roll back an AI coding agent's changes to a known-good point.
- Claude CodeClaude Code is Anthropic's official command-line agent that plans, edits, runs, and verifies code across your repo using Claude models and tool use.