close
Skip to content
Bring your own agent

The harness for engineers who care about design.

Your repo and coding agents, right beside Wonder’s design canvas. Work on parallel tasks, run your app next to the design and take each change through review.

BERJAYA
BERJAYA
BERJAYA

Used by designers & engineers at:

  • Autentic
    BERJAYA
  • Deeploi
    BERJAYA
  • Arc
    BERJAYA
  • Monaco
    BERJAYA
    BERJAYA
    BERJAYA
  • Browserbase
    BERJAYA
  • Notion
    BERJAYA
  • CodeRabbit
    BERJAYA
  • Tenderly
    BERJAYA
  • Steel
    BERJAYA
  • Preview
    BERJAYA

Your repo. Your agents.
Room to work.

Parallel agents that don’t collide. Each workspace is its own git worktree, a separate copy of your branch.

Team handoff. Open a teammate’s branch or PR as a workspace and build on their work.

Your subscriptions.
All in one place.

Opus 5.5 selected with Claude Code

Your agents, shared context. Run Claude Code and Codex on your subscriptions, share chat history and switch models.

Manage multiple accounts and choose the Claude Code or Codex subscription, model and effort for each chat.

Run it. See it. Shape it.

Run your app in one click after setting up its run command, and open the Preview beside your agent chat.

From branch
to merge.

Git & GitHub support built in. Create branches, commit, push and open pull requests without leaving the app.

Review before you merge. Get a second agent’s review and catch issues early.

Planned

Cloud workspaces. Keep agents working with your laptop closed.

Frequently asked
questions.

Something else on your mind? Talk to the team.

Why use Wonder over other coding agents?

Each of those runs only its own agent. To run several tasks at once you’d have to create git worktrees, install dependencies in each one and switch between terminal windows yourself. Wonder does that setup for you. It shows every agent in one window, lets you mix Claude Code and Codex in the same project and puts a design canvas next to your code.

How does the canvas fit in?

It’s the heart of Wonder, where you see and shape what your agents build. Designs on the canvas are real front-end structure, not pictures, so work moves both ways. An agent brings a screen from your code onto the canvas, you change it there, and the agent builds the change into your code. Your running app also opens on the canvas as a live Preview.

Does Wonder support running agents in the cloud?

Not yet. Today agents run on your own machine. We’re working on Wonder Cloud so you can run them in the cloud too.

How does Wonder compare to other IDEs?

Compare them on what Wonder is built around: an isolated worktree for each task, several coding agents in one app, a built-in Git and GitHub workflow, and a design canvas next to your code.

Do I need to know Git and GitHub?

For code, yes, at least the basics. Wonder’s agent chats are built around Git and GitHub. Every worktree is its own branch. You commit and push from the app, and you follow each pull request’s CI checks and reviews there too.

Do I need my own AI subscription?

Yes. Wonder is the interface, and the agents run on the Claude Code or Codex subscription you already have.