Project tour and code map: get to know a codebase fast
When you open a new repository, or wonder where your agents have been working, two tools on the project details page help: a step-by-step project tour and a map of the code.
Tell me about this project
The tour covers a summary, the stack, modules, key files, a flow diagram whose boxes open the related file, first steps, package.json scripts and things you should know. The window notes when the project has changed since the tour was made.
- The tour is prepared by a headless Claude Code run (claude -p, Sonnet by default).
- The agent can only use Read, Glob and Grep: no edits, no commands, no tool servers.
- File contents it reads go to Anthropic through your own Claude Code account, as with any Claude agent.
Code map
A module dependency graph for TS/JS, Python and Go, with a folder depth from 1 to 4. The “most touched files” and “most broken” lists are computed locally from the latest 80 Claude Code transcripts on your computer. “Broken” means a command failed later in the turn in which the agent edited the file.
Subagents
Lists the subagents a Claude Code agent started, with type, whether it ran in the background, status, tool count and duration, plus a timeline of each one’s steps. It reads local transcripts and works with local Claude Code agents for now.
Questions
Does the tour send my code anywhere?
The tour is made with Claude Code, so the content of the files it reads goes to Anthropic through your Claude account. The code map and subagents view are fully local.
Can the agent change files while making the tour?
No. Only Read, Glob and Grep are allowed; editing and running commands are off.
Which languages appear on the code map?
TS/JS, Python and Go.
Is the subagents view available for Codex?
No, for now it works with local Claude Code agents only.
Related features
Bring your agents to one desk.
Start on the free plan. Download it, add a project folder and open your first agent.