MCP
Summary
- MCP (Model Context Protocol) is a standard way for AI clients to connect to external tools and data sources.
Key Points
- MCP provides a structured interface so models can discover and call external capabilities.
- Shared concept across ecosystems: expose tools/resources to agents through a protocol layer.
- Practical benefit: one integration can serve multiple clients if they support MCP.
- Differences still exist by client: auth flow, server startup, permissions, and UI exposure.
- Security matters: restrict tool scope, protect secrets, and audit what tools can execute.
Examples
- Connect a local MCP server that exposes filesystem search tools.
- Add an MCP server for internal docs so the agent can query team knowledge.
- Use MCP to unify access to issue tracker, database read APIs, and deployment metadata.
- AI/General/Skills
- AI/General/Sub-agents
- AI/General/Hooks
- AI/OpenAI/Slash Commands Reference
- AI/Claude Code/Slash Commands Reference