Every line of code was written for a reason.

Ask why a piece of code is the way it is. Repo Archaeologist follows it to the commit that changed it, the pull request behind that commit and the issue behind the pull request, and cites each one.

  1. Code now

    F-01Markup.striptags in src/markupsafe/__init__.py__init__.py:199–238
  2. Commits · what changed

    Nothing at this depth

  3. Pull requests · the goal

    F-03Oct 2026fix performance regression in striptags#522F-02Jan 2024improve striptags performance#413
  4. Issues · the need

    F-04Apr 2026Performance regression in striptags#521
A real run on pallets/markupsafe, showing the evidence its answer cites: “Why does Markup.striptags collect the text between tags into a list in a single pass, instead of removing tags from the string one at a time?”Watch it in full

git blame tells you which commit. It rarely tells you why.

The reason is usually a hop or two away: in the pull request that commit came from, or the issue that pull request fixed.

  1. Plan

    Reads the question and names the code it is about.

  2. Search

    Code, docs, commit messages and pull requests, at once.

  3. Follow the links

    Commit → pull request → issue, one hop at a time.

  4. Write and check

    Every claim cites a find, and every claim is checked.

What people ask it

Getting started in a codebase
“What does this module do, and where does a request enter it?”
An explanation of the code as it is now, citing the files and functions it read.
Reviewing a change
“Why is this check here?”
The commit that added it and the discussion that asked for it.
Refactoring
“Is this workaround still needed?”
The issue it worked around, so you can see whether that problem still exists.
Debugging
“When did this behaviour start, and why?”
The change that introduced it, and the pull request that explains the intent.

Real runs, replayed

All examples

Cited, checked, and yours to run

Every claim cites a find

Answers point at the commit, pull request, issue or line of code behind each sentence, and a second pass checks every claim against what it cites. In the examples here, 8 of 8 claims were supported by their evidence.

The source is public

The agent, the website and the evaluation are on GitHub. Read how it decides what to follow before you trust it.

Run it yourself

One script on your machine, or one Docker image on a server. A Render blueprint and a Vercel setup are included. How to deploy it.

Free to start

It runs on Gemini's free tier, slowly, or on Claude. Your API key stays on your server.

Pick a repository. Ask why.