Skip to content

Where people and agents ship software together. The open-source git platform for the whole job: issues, agents, checks and deploys to the edge.

g1t.sh
syntaqxA workspace's owners get Getting started, a setup guide that opens by itself on Home and the Workspace page until they finish it or close it. It is a dialog with a rail of six steps beside the step at hand (on a phone, the sheet with "Step n of 6" and a progress bar above it): add an AI provider, choose Ask's model, bring your code, connect your harness, invite your team, add a runner. Each step is ticked from what the workspace has, never from clicks (lib/onboarding.ts, read by lib/onboarding.server.ts at once from integrations, projects, the GitHub App, identity, billing and actions): a provider whose key passed its last check; Ask answering with a provider and no problem; a project, or GitHub connected; a token named "g1t CLI on …" or an app signed in with g1t, as the Claude Code plugin does; another member or an invitation waiting; a runner. The AI steps work in place: choosing a lab (or a platform or endpoint from More providers), pasting its key and Add and check the key, Test, and Use for Ask all post to the AI page's own action, so they are the AI page's checks and the page reads the new state after. Bring your code offers New project and the one-click Connect GitHub; the harness step has the CLI's install commands for macOS and Linux and for Windows, g1t auth login and the plugin's two commands, ready to copy; on a free workspace the team step says workspace invites come with the plan, links to it, and offers inviting a friend to g1t instead. Every step but the provider can be skipped, and when everything is done or skipped it says All set and Finish puts it away. AI stays a base setting: nothing of it is in Integrations.