OpenWorkProof MCP server

OpenWorkProof is a helper that keeps a signed record of the work an AI does for you. It writes down who allowed the task, what the AI was allowed to touch, what it actually did, and whether a person accepted the result. It is handy if you let an AI change files or run commands and you want proof later of what happened.

What this MCP server does

You ask your AI to do a task, like changing a file or running a test. Before it acts, the helper checks whether that action is allowed under the rules you set. If it is allowed, the AI goes ahead and the helper writes down a signed note of what it did. If the action is one a person must approve first, the helper stops it and asks for a human decision. At the end you can export a small evidence pack that someone else can check without logging into your systems.

Click to zoom

What you can do with it

• Record who authorized a piece of AI work
• Check whether an action is allowed before the AI runs it
• Pause actions that need a human decision first
• Sign a note of each important action the AI takes
• Verify that a chain of evidence is complete and unchanged
• Export an evidence pack for someone else to check offline
• Keep verification and final human acceptance as two separate steps

What it gives back to you

You get back short answers in the chat: whether an action is allowed, reserved for a human, or denied. You can also get a list of the recorded steps, a summary of a verification result, and a folder of signed files you can save or send to someone. The files can be checked later by anyone who has the pack and the matching public key.

Before you start

Who it's for

Teams and people who let an AI change code or run commands and want a signed record they can show someone else later.