Model Context Protocol (MCP)
Model Context Protocol is the open standard introduced by Anthropic in late 2024 for connecting LLM applications to external data sources and tools. MCP defines a uniform schema for declaring tools, resources, and prompts so that any compliant client (Claude Desktop, Cursor, custom agents) can use any compliant server (databases, file systems, APIs).
MCP addresses the M×N integration problem: every LLM client used to need a custom integration for every data source. With MCP, both sides implement the protocol once and interoperate. The protocol is straightforward — JSON-RPC over stdio or HTTP, with three primitives (tools, resources, prompts) — but its impact is structural: the ecosystem of MCP servers (GitHub, Postgres, Slack, file systems, browser automation) is the agent equivalent of npm packages. Adoption accelerated through 2025-2026 as Anthropic, OpenAI, Google, and other AI vendors converged on MCP as the shared standard.
Related terms
- Tool use (LLM agent)
Tool use is the LLM-agent pattern in which the model has access to a defined set of tools — read file, search web, run code, query API — and decides which to invoke and with what arguments based on the user's request and the current state.
- Function calling
Function calling lets an LLM invoke external tools by emitting a structured request matching a function specification provided in the prompt.
- Agentic workflow
An agentic workflow is a software process in which one or more LLM-driven agents drive the work, making decisions and invoking tools at each step rather than following a hardcoded script.