What is the Model Context Protocol (MCP)?
The Model Context Protocol (MCP) is an open standard that lets AI agents call external tools, services and data sources through one common interface. In short: MCP is what turns a chat model into an assistant that can actually do things — read your files, run a Git command, query a database — instead of only describing how.
Why MCP exists
Before MCP, every AI tool integration was bespoke: each assistant wired up each tool in its own way, and none of it was portable. MCP replaces that with a shared protocol. A model that speaks MCP can connect to any MCP-compatible tool server, and a tool exposed over MCP can be used by any MCP-capable client. The result is fewer one-off plugins and far less lock-in.
How MCP works, briefly
An MCP setup has two sides: clients (the AI application, such as an IDE) and servers (the tools and data sources). The server advertises what it can do — functions it exposes, resources it can read — and the agent calls those capabilities as needed during a task. Because the contract is standardised, the same server works across different models and apps.
What it unlocks for developers
With MCP, an agent can open and edit files, run version-control actions, hit an internal API, or pull context from documentation — all as steps in a single workflow. That moves AI from suggestion to action, with you approving what runs.
Is MCP tied to one vendor?
No. MCP is an open standard, which is the point — the tools you connect aren't locked to a single model or company. Pair MCP with whichever model you prefer; in Eclipse you can bring any LLM and keep your tool integrations when you switch.
FAQ
Is MCP only for coding tools?
No. MCP is general-purpose, but it is especially useful in development tools because agents can act on code, version control and project data. Any system that exposes an MCP server can be connected.
Do I need MCP to use AI in Eclipse?
No. You can use plain AI chat with any connected model. MCP is what you add when you want agents to take actions through real tools rather than only generating text.