MCP Evals — can an agent actually use your server?

A passing connection test proves your server speaks the protocol. It does not prove that a model handed your tools can work out which one to call, or that the answer it comes back with is right. An eval is a real task — “what is the total for order 1041?” — given to a model that has nothing but your tool names and descriptions to go on.

The most valuable thing it finds is the failure no schema check can see: every tool call succeeded, the arguments validated, and the answer was still wrong. Below is a finished run on a made-up orders server, replayed exactly as the product renders it.

DemoA finished run on a made-up orders server, replayed. Nothing connects to a real server, no model is called, and nothing is charged.

Detects your tools, writes real tasks against them, then hands each task to a model that has to complete it using your server. Tells you whether your server is actually usable — and which models can use it.

Run this against your own server

Point it at any MCP server you can reach. It detects the tools, harvests real values from the read-only ones so the tasks are grounded in your actual data, writes the evals, then runs them. Tools that write are left out unless you opt in.

Run evals on your server

Questions people ask

What is an MCP eval?
A task a model has to complete using only your server’s tools. It checks whether an agent can pick the right tool from your descriptions and reach a correct answer — not whether the endpoint returns 200.
How is this different from an MCP client or inspector?
A client tells you the protocol works. An eval tells you the server is usable. If a tool description omits that a query also matches email, an inspector will never notice; an eval shows the agent burning five calls to find one customer. Try the inspector if the protocol is what you want to check.
Are the results deterministic?
No. Models write, drive and grade the evals, so the same suite can come out differently next time — and we only ever see what your server exposes, never your database or what the right answer should have been. Treat a run as evidence, not proof.
Do I need a server to try it?
Not for the walkthrough above — it is canned. To run evals for real you need an MCP server URL we can reach, or one of the free public mock servers.
MCP Evals — Test Whether an AI Agent Can Actually Use Your MCP Server | MCP Playground