Y Combinator-backed Onlook launches open-source visual canvas for coding with AI agents
Onlook, part of YC's W25 batch, has launched an open-source desktop tool that lets designers visually edit a live, running version of their actual codebase rather than a mockup. The tool doesn't include its own AI model; instead, users connect existing Claude Code or ChatGPT/Codex accounts, which write the code while designers shape the interface on a canvas. It works directly with local Git repositories and dev servers, and supports MCP connectors for additional context.
GoKawiil's interpretation of the reporting above, not reported fact.
By skipping its own AI model and plugging into agents users already pay for, Onlook could lower the barrier for designers to generate production-ready code without engineers rebuilding their mockups. Its open-source, local-first approach contrasts with closed, cloud-hosted rivals like Figma, V0, and Webflow, potentially appealing to developers wary of vendor lock-in or data leaving their machines. Whether this bring-your-own-agent model can match the polish of integrated AI design tools remains to be seen.
- Onlook is an open-source, desktop-based visual canvas for editing live code, launched via Y Combinator's W25 batch.
- It relies on external AI agents (Claude Code, ChatGPT/Codex) rather than a proprietary model, with MCP connector support.
- It edits a user's real Git repository and running dev server directly, aiming to eliminate the mockup-to-code handoff.
Source: github.com, 2026-11-19
Published there as: “Launch HN: Onlook (YC W25) – An open-source developer tool for designers”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.