WhyGraph¶
Explains why code exists, not just what it does.
A rationale layer over CodeGraph. It mines your git history and GitHub for the story behind each line - the commits, pull requests, and issues that put it there - and serves that story to any AI editor over MCP, or to you in the WhyGraph portal.
-
Evidence from your history
For any chunk of code, WhyGraph pulls the commits that touched it, the blame behind each line, the PRs that merged it, and the issues those PRs closed - already linked together.
-
Rationale cards
Ask why a symbol exists and get a structured card: purpose, why, constraints, tradeoffs, and risks. Each card is cached, so the second lookup is instant.
-
MCP-native
Every project gets a standard HTTP MCP endpoint on the portal. Claude Code, Cursor, VS Code, Codex - any editor that speaks MCP can call it, and the portal wires each one for you.
-
Only Docker required
No Python, Node,
gh, or CodeGraph on your host. A tiny shim runs everything inside one image. Install,whygraph up, open the browser - done. -
The portal
whygraph upstarts one local server for all your projects: an Explorer over the code graph with rationale and evidence side by side, plus a chat assistant that answers questions by calling the same tools. -
Composes with CodeGraph
CodeGraph answers "what's connected to what". WhyGraph answers "why it exists and when it changed". The portal runs CodeGraph's own MCP server for you, beside WhyGraph's; each stays focused on its own job.
-
Git analysis as a service
The MCP endpoints aren't just for editors. Real applications can connect to them for git-based analysis of a target repo, reading the same cached data your scans write.
Who it's for¶
You're dropping into an unfamiliar codebase, or editing code you wrote months ago and no longer remember. The what is in front of you; the why is buried in history. WhyGraph surfaces that why right where your AI assistant works, so an edit respects the original intent instead of rediscovering it the hard way.
Ready? Start with the Quickstart.