DiffGate MCP server

DiffGate is a small helper that checks the code your AI assistant writes before that code is saved to your computer. It looks only at the lines that changed and flags risky things like hardcoded passwords, unsafe database queries, or edits to login code. It is handy for developers and anyone who lets an AI write or edit code for them.

What this MCP server does

You ask your AI to write or change some code. Before saving, the AI sends the changed lines to DiffGate. DiffGate compares them to the last saved version and checks those lines for common problems, like passwords left in the code, unsafe database queries, or changes to login and payment logic. It sorts findings into three levels: green (safe to keep), yellow (take a quick look), and orange (check carefully before keeping). The AI then shows you what it found and can fix the problem right away, while it still remembers what it was doing.

Click to zoom

What you can do with it

• Check AI-written code changes for security problems before they are saved
• Spot hardcoded passwords, keys, and tokens in changed lines
• Flag unsafe database queries and command calls
• Warn when login, payment, or database schema code is touched
• Sort findings into green, yellow, and orange so you know what to look at
• Show what the AI changed and why, instead of silently rewriting
• Run your project's tests only when a change is high-impact

Try asking your AI

• “Write a login endpoint for my Express app and check it with DiffGate before saving”
• “Review the changes you just made to the database file and tell me if anything is risky”
• “Add a file upload feature and make sure DiffGate does not flag any secrets”
• “Check my last few commits for anything that needs a closer look”

What it gives back to you

You get back a short list of findings, each tied to the exact lines that changed. Every finding has a level (green, yellow, or orange), a plain reason, and often a suggested fix. If the AI fixed something, it shows you the original line, the new line, and why it changed. You also see the result of your tests when a high-impact change triggered them.

Before you start

Who it's for

Developers and technical teams who let an AI assistant write or edit code and want a quick safety check before those changes land.