The Model Context Protocol, often shortened to MCP, is an open standard for connecting AI assistants and agents to outside tools and data. These connectors are small programs called servers. Each one gives the assistant access to one thing, such as a calendar, a shared drive or a database. Every server speaks the same protocol, so one assistant can use many tools in the same way. Think of it like a universal plug: your donor database and your email inbox could both connect to the same assistant without custom work for each.
Why it matters
Every connector you add gives the assistant more access. Every connector is also a piece of software that can have security flaws. A connector to your shared files lets the assistant read those files, so treat installing one like handing out a key. Before you add a connector, ask three questions: who made it, whether it is still maintained, and exactly what it can see or change. Install only connectors you trust, keep them updated, and remove the ones you no longer use. If an assistant can act through many connectors, read about excessive agency. You can also check our advisories for known problems.
Published Updated