sofagent audit

sofagent is a helper that watches over your AI coding assistant. Every time the AI changes code, sofagent looks at what changed and warns you if something looks wrong, like a password left in the open or an edit outside the agreed scope. It is handy if you let an AI write or change code for you and want a record of what it did.

What this MCP server does

You ask your AI to check the latest code change. The AI passes that request to the sofagent helper. The helper looks at the difference between the old and new code, checks it against a set of safety rules, and saves a record. Then it sends back what it found, so you can see if anything needs fixing.

Click to zoom

What you can do with it

• Scan the most recent code change for leaked passwords or keys
• Check whether an edit stayed inside the agreed scope
• Look for signs of prompt injection in commit messages
• Save a tamper-evident record of each audit
• Roll back to an earlier snapshot if something went wrong
• Show a summary of past audits and violations

Try asking your AI

• “Check the last commit for any leaked secrets”
• “Did my AI edit files outside the project scope?”
• “Show me the audit history for this repository”
• “Roll back to the snapshot before the last change”

What it gives back to you

You get back a short report in the chat. It tells you if the change passed or failed, and lists any problems it found, like a leaked key or an edit that went too far. It can also show a history of past audits or confirm that a snapshot was restored.

Before you start

Who it's for

Anyone who lets an AI write or change code and wants a simple safety check and a record of what happened.