Agent concepts
MCP server
An MCP server is a program that exposes tools, resources, and prompts to AI agents over the Model Context Protocol.
Definition
What is an MCP server?
An MCP server is the provider side of the Model Context Protocol. It advertises a set of tools (actions the agent can invoke), resources (data the agent can read), and prompts (templated instructions), each with a name and a natural-language description the model uses to decide when to call it. Servers run locally over stdio or remotely over HTTP.
Because an agent reads and acts on the text an MCP server sends, the server sits inside the agent's trust boundary. A server can request more capability than it needs, describe a tool in a way that hijacks the model, impersonate another server's tools, or quietly change its definitions after being approved. Connecting to an MCP server is a trust decision that deserves the same scrutiny as adding a dependency.
FAQ
Common questions.
How do I know if an MCP server is safe to connect?
Inspect the tools it declares before connecting, prefer servers with a known publisher and provenance, watch for tools that request dangerous capabilities or solicit credentials, and monitor for changes to the definitions after approval. A security gateway can score a server on these signals automatically.
Related terms
Keep reading.
Govern the actions, not just the vocabulary
AxioRank scores every tool call your agents make for leaked secrets, PII, destructive operations, and prompt injection, checks it against your policy, and proves it in a tamper-evident audit log. Start free, no card.