Quickstart¶
You've installed WhyGraph. The happy path is three moves: install, whygraph up,
browser. There is no per-repo command to run first.
1. Start the portal¶
This starts the WhyGraph portal in the background, 127.0.0.1 only, with its database in a second
container, and shares that folder with it. Drop --add-folder to start without sharing anything yet; you can share later. See
Start the portal and Shared folders.
2. Open it in the browser¶
Open http://127.0.0.1:8765. The first time, enter your name on the Welcome screen. You land on an Projects page, which starts with a short checklist: add an LLM key, add a project, connect your agent. Each step ticks itself off; Dismiss hides the list and Getting started brings it back.
3. Add a project¶
Choose New project, pick a repository from your shared folders, and follow the three steps:
- Source - the repository.
- Set up - review the preview and choose which agents to connect. Set provider keys once under Settings and every project inherits them.
- Configure - the first scan starts on its own: it reads git history and refreshes the CodeGraph index, with live progress, and it never calls an LLM, so it is free and fast. This step also shows the estimate for the optional descriptions.
Enter your keys in the portal
ANTHROPIC_API_KEY, GH_TOKEN and the other credentials in your shell environment do not
reach WhyGraph. Enter them under Settings. See Upgrading from 1.x.
Adding projects covers each step in detail.
4. Browse, then ask your editor¶
The project opens in the Explorer: the code graph with rationale and evidence side by side, and a Chat assistant that answers questions by calling WhyGraph's tools.
Your agent connected during Set up. Approve the whygraph server when it asks (Claude Code) or trust
the project (Codex), then ask why a function exists and WhyGraph answers from history. See
Connecting agents.
Descriptions when you want them¶
After the first scan, the project offers Describe core and Describe all, each with its commit count, the model and a cost estimate, to write the LLM description for each commit. Core covers the commits behind the files the rest of your code depends on; all covers every commit. Skip both and descriptions backfill on demand. What you spend, per project and model, is tracked under Usage & cost. See Scanning your repo.