EvalView MCP server

EvalView is a tool that records what your AI agent does today, then tells you when that behavior changes after an update. It is handy for anyone building or maintaining an AI agent who wants to catch surprises before users do. If you have ever changed one line and wondered what else broke, this is for you.

What this MCP server does

You ask your AI to snapshot your agent, and it runs EvalView to record the tools your agent calls, in what order, and with what output. Later, after you change something, you ask it to check, and EvalView compares the new behavior to the saved baseline. It tells you if the tools changed, the order changed, or the output quality dropped. You get a clear pass, warning, or fail for each test, right in the chat.

Click to zoom

What you can do with it

• Record a baseline snapshot of your agent's current behavior
• Check your agent against the baseline after a change
• See which tools were called differently or in a different order
• Spot when output quality drops compared to the baseline
• Accept a new behavior as the updated baseline when it is correct
• Run a quick demo to see how it works without setting up your own agent

Try asking your AI

• “Take a snapshot of my agent's behavior so we have a baseline”
• “Check my agent against the baseline and tell me what changed”
• “Did the refund flow change after my last edit?”
• “Show me the diff for the billing dispute test”

What it gives back to you

You get a short report for each test: passed, tools changed, or regression. It shows which tools were called differently and in what order, plus any drop in output quality. If you asked for a diff, you see the before and after side by side. The whole thing appears as a simple list in your chat.

Before you start

Who it's for

Developers and testers who build or maintain AI agents and want to catch behavior changes before they reach users.