How to understand a new codebase fast

Find where work comes in: HTTP routes, CLI commands, queue workers, the main function. Then pick the most important user action (sign in, checkout, a request) and follow it call by call until a response goes out. One traced flow teaches more than an hour of browsing folders.

These map most codebases quickly:

• Where is routing handled, and where is configuration loaded?
• How does a request get handled, step by step?
• Which tests cover the part you will change first?
• What changed recently, and why (the pull requests, not just the diffs)?
• What is unused, so you can ignore it?

Synapse maps any public GitHub repository into a code graph you can ask questions of, free and without signing up: paste the repository at asksynapse.in/github. Ask “how does a request get handled?” and it traces the real call path through the code, each step linked to the file and lines, with an AI summary written only from the traced code. It also answers where a feature lives, which tests cover it, and what looks unused.

For private repositories, the same memory reaches Claude Code, Cursor and Claude over MCP, so your agent starts a task already knowing the codebase.

Synapse is free to start: connect a GitHub repository and add it to Claude Code, Cursor or Claude in about two minutes. Create an account or read the docs.