QuestionsWhat an AI agent editor IDE is, and what Yiru does with the idea.

One worktree per task

Two agents editing the same checkout overwrite each other, and a half-finished change makes the next task start from a dirty tree. Yiru gives each task its own git worktree, with its own files, terminal state and history, so several agents can work at once and you compare the results before anything merges.

Thirty-five agents, as themselves

Claude Code, Codex, Gemini CLI, Copilot CLI, Cursor, Aider, Amp, OpenCode, Goose, Cline, Continue, Factory Droid, Grok and twenty-two more run as the terminal programs they already are. Yiru starts, monitors and resumes their sessions and keeps provider-specific behaviour isolated, so your authentication, model access and usage limits stay where they were.

Wherever the code actually lives

A workspace runs on your own machine, inside a WSL distribution, over SSH, or through a relay-connected host. Filesystem, git, terminal and search operations are routed to whichever host owns the worktree, and cached host state never leaks between them — so a repository that only exists on a build box is still a normal workspace.

Review the evidence, then merge

Read the diff, annotate individual lines to queue follow-up work, inspect checks, and open or read pull requests on GitHub, GitLab, Azure DevOps and Gitea. The iOS and Android companions pair with the desktop app for the same review and merge, for the part of the loop that happens away from the desk.

What is an AI agent editor IDE?

It is a workspace built around coding agents rather than around a text cursor. The editor still shows files and diffs, but the unit of work is an agent session: you give it a task, it works in its own checkout, and you review the result. Yiru adds the part a chat window cannot — somewhere for several of those sessions to run at once without colliding.

Which coding agents does Yiru run?

Thirty-five, including Claude Code, Codex, Gemini CLI, Copilot CLI, Cursor, Aider, Amp, OpenCode, Goose, Cline, Continue, Factory Droid and Grok. They run as the terminal programs they already are, on whichever host you point Yiru at, so authentication, model access and usage limits stay with the agent provider.

How is this different from an AI code editor like Cursor?

An AI code editor puts a model inside one editing session. Yiru runs the agents you already use, several at a time, each in its own git worktree, and gives you the review surface for the diffs that come back. It does not replace your editor or your agent — it is where they run.

Why isolated git worktrees?

Two agents editing one checkout overwrite each other, and a half-finished change makes the next task start from a dirty tree. A worktree per task gives each agent its own files, terminal state and history, so you can run tasks in parallel and compare the results before anything merges.

Can it work on a remote machine?

Yes. A workspace can execute on your local machine, inside a WSL distribution, over SSH, or through a relay-connected host. Filesystem, git, terminal and search operations are routed to whichever host owns the worktree, so the code never has to live on the laptop.

What can the mobile app do?

The iOS and Android companions pair with the desktop app to watch sessions, take notifications, read the diff, send a follow-up instruction, and merge. It is for the part of the loop that happens away from the desk, not a second editor.

Does it review pull requests?

It reads and creates them on GitHub, GitLab, Azure DevOps and Gitea, including discussions, checks and review state, and it can annotate individual diff lines to queue follow-up work for an agent.

Is Yiru free?

Yes, and open source under the MIT licence. Desktop builds for macOS, Windows and Linux come from GitHub releases; the iOS app is in TestFlight and Android ships as an APK. You pay whoever provides your agents, not us.