Wonder
An AI agent harness and playground for design and code
927 followers
An AI agent harness and playground for design and code
927 followers
Wonder is an AI agent harness and a playground for building products. Run Claude Code and Codex in parallel, manage multiple subscriptions, and review their work in one place. Bring your own context and workflows, explore ideas on the canvas, and turn them into working software, all in one workspace.
This is the 2nd launch from Wonder. View more
Wonder Dev View
Launching today
Run Claude Code and Codex on your subscriptions, in parallel worktrees. Have one review the other’s code, then try your app beside the design. Select an element in Preview and tell your agent what to change.
Free Options
Launch Team / Built With
AWS Activate CreditsCredits that fuel your startup’s next milestone
Promoted
Hi, I’m Aibek, co-founder of Wonder.
We’re excited to launch Wonder Dev View - a harness where you can orchestrate multiple Claude Code and Codex agents on your repos.
Wonder is built locally, so it connects your agents, MCPs, skills and rules automatically. Keep each task’s chats, changes, terminal and pull request together. You can choose a different model to review the code, including Claude reviewing Codex or the other way around.
The added bonus is a powerful canvas to visualize your designs, system diagrams and anything else. Visualize multiple versions of your designs, branch out and view multiple previews side by side on the canvas. Winder Wonder, you get a lot more control over your designs, get better output and work directly with designers in one tool.
For code-only tasks, stay in the full-screen coding view. Multiple accounts and optional saved-memory sharing between Claude Code and Codex help you keep working across providers.
Wonder runs on Mac and Windows. It’s free to start; shared canvas editing and publishing a Preview link are on Pro. Your agent subscription stays separate.
Try it on one task in your existing repo! Wonder Cloud coming soon. What would you need before making it your daily setup?
@superaibek Excited to launch this with you, Aibek!
Been wanting something like this. Switching between Claude Code and Codex gets messy, so having both work side by side is actually pretty cool. Congrats on the launch! Nice hunt @thisiskp_
@thisiskp_ @trulysurya Thanks, Surya! You can keep both in the same workspace and choose which model handles each task. Curious how you split work between Claude Code and Codex today
if Claude code and codex work together side by side will it hallucinate the same error or response stays separate ?
Keeping the chat, diff, terminal and pull request together gives a reviewer useful context. I'd also want to see whether the other agent reviewed the current diff or an earlier version of it.
@oleksii_sekundant The review card shows the commit it reviewed and marks it stale when the branch moves to another commit. That check is commit based, though, so it doesn’t catch every uncommitted edit. If the code changes after a review, I’d run it again
Hey PH community 👋
Excited to hunt Wonder Dev View today and support my friend Aibek!
Wonder brings Claude Code and Codex into one place, with each task’s chats, code changes and previews kept together. You can have one model write the code and another review it, then compare design options side by side on the canvas.
Give it a spin on an existing repo and let Aibek know what you think!
@thisiskp_ Thanks for hunting us, KP! Really appreciate you helping us get Wonder in front of more developers
I built my own product alone with Claude Code as the only teammate, so "have one model review the other's code" speaks to me. The review card that names the commit it looked at and marks itself stale when the branch moves is the detail that makes the review trustworthy. Do you plan to let the reviewer run the tests in its read-only context, or is that left to the author agent on purpose?