DiffContext MCP server

DiffContext is a helper for Python codebases. You tell it what you are about to change, and it finds the small set of functions your AI actually needs to see: the ones that call your function, the ones that override it, the tests that cover it. It is handy if you work in a big Python project and your AI keeps missing parts of the code.

What this MCP server does

You ask your AI to change something in your Python project. The AI asks DiffContext which functions are connected to that change. DiffContext looks at the project's structure and finds the callers, the subclasses and the related tests. It then packs the most useful ones into a size that fits your AI's limit, and tells the AI what it had to leave out. Your AI gets a small, focused bundle of code instead of the whole repository.

Click to zoom

What you can do with it

• Find the functions that call the one you are changing
• Find subclasses that override a method
• Find tests that cover a function
• Pack the relevant code into a size your AI can handle
• Report what was left out of the bundle
• Check how well the selection matches your project's own history

Try asking your AI

• “I need to change validate_jwt in src/auth.py. What else in the project touches it?”
• “Show me the callers and tests for the function I am about to edit.”
• “Give me a small bundle of code around this change, under 8000 tokens.”
• “Which subclasses override this method, and will my change break them?”

What it gives back to you

You get a short bundle of code in the chat, ready to read or paste. It usually includes the function you are changing, the functions that call it, related subclasses and a few tests. There is also a short note at the top saying what was left out, so you know what the AI did not see. Sometimes it will say it could not find anything relevant, which is also useful to know.

Before you start

Who it's for

Developers and reviewers who work on large Python projects and use an AI assistant to help make or review changes.