Agent Mode 2.0
Give the agent a goal. It explores your project, edits a private draft, runs your project’s checks on the device, and presents a patch for you to review before anything changes. Works with Claude, OpenAI, Gemini, Grok or OpenRouter using your own key.
Tasks that survive interruptions
Agent tasks save checkpoints you can review or resume, and can work on their own Git branch. Cloud work can continue after you switch apps where the system allows it; checks that run code on your device wait until you return.
Everyday Git is free
Clone over HTTPS or SSH, see status, stage, commit, resolve conflicts, pull, push, sync and view diffs without Pro. Pro adds branches, rebase, tags, history, blame, stashes, pull requests and AI commit messages and diff reviews.
Pull and merge requests, end to end
See CI checks and logs, check out a branch to run it locally, review with inline comments and suggested changes, approve, and merge exactly the commits you reviewed. Works with GitHub, Forgejo/Gitea and, new in 3.0, GitLab.
From a failing check or an issue
Fix with Agent starts from a failing CI check, with its log, and Start Work turns an issue into a branch you can hand to the agent.
Navigate code from the editor
Go to Definition, Find References, Rename, Show Documentation and Quick Fix from the keyboard, the context menu or a modifier-click, for supported JavaScript, TypeScript and Python projects.
Dependencies, SQLite and async debugging
Install pure npm packages and Python wheels for offline use, run SQL scripts with SQLite, and step through async JavaScript and TypeScript with breakpoints and locals.
Sign in with Grok
Grok can use an xAI API key or, with an eligible Grok account, Sign in with Grok. Plans and models are controlled by xAI.