- move faster on real repositories (not toy demos),
- run multiple specialist agents at once,
- keep tool runs and changes observable,
- and reduce the “what just happened?” moments that usually come with agent automation.
Start here
If this is your first time setting up the app, start with Quickstart. If you already have the app installed, the fastest way to get value is:- Open a workspace (a folder with code you care about).
- Create a new thread.
- Pick a provider + model.
- Ask for a small, concrete change (one file, one feature, one fix).
- Watch the timeline as the agent uses tools, and approve what you’re comfortable with.
Key concepts (mental model)
- Workspace: the folder you opened (typically a repo). gg keeps sessions and file views scoped to it.
- Thread: one agent chat session. Think “one line of work” with its own context and timeline.
- Team: a group of threads working together. One thread is the lead; the rest are members.
- Tool calls: actions an agent can take (run commands, edit files, send messages). gg surfaces these so you can review what happened.
- Background processes: long-running commands that keep going without blocking the conversation, with results injected back into the owning thread.
Documentation map
Use these sections as your primary entry points:- Getting started for install and first-run setup.
- Guides for practical workflows and delegation patterns.
- How it works for a high-level view of the runtime, tools, and teams.
- Auth for provider-specific authentication setup.
- Configure for preferences, policies, and customization.