RealRun MCP server

This is a small helper that lets your AI look up independent test results for AI models on everyday work tasks. RealRun runs these tests and publishes the scores. It is handy if you are choosing a model for a job like sorting email or handling receipts and want real numbers instead of marketing claims.

What this MCP server does

You ask your AI something like which model is best for sorting email. Your AI sends that question to this helper. The helper looks up RealRun's published test results for that task. It sends back scores, a recommended model, and any safety warnings. Your AI then explains the answer to you in the chat.

What you can do with it

• See which everyday tasks RealRun has tested
• Get every model's score for a task, best first
• Ask for a model recommendation based on accuracy, cost, or a balance of both
• Check safety warnings for a task, like scam failures or invented data
• Compare models before you pick one for a job

Try asking your AI

• “Which AI model is best for sorting my inbox?”
• “Show me the scores for meeting scheduling.”
• “Recommend a model for receipts if I care most about cost.”
• “Are there any warnings for the support replies task?”

What it gives back to you

You get back plain answers in the chat: a list of tested tasks, a ranked list of model scores, or a single recommended model with two alternatives. Warnings come back as short flags, like a model that failed on scam messages. Everything is read-only, so nothing in your accounts changes.

Before you start

Who it's for

Anyone picking an AI model for everyday office tasks, like operations staff, small business owners, or support team leads.