Google Labs introduces Stitch Loop product agent for shipped code
The experimental agent connects to a team's GitHub code and live product data, proposes fixes as pull requests, and opens to a global waitlist first.

- What happened
- Google Labs announced Stitch Loop, an experimental agent that keeps working on a product's code after launch.
- What’s new
- Unlike coding agents that stop once a pull request merges, Stitch Loop runs continuous loops toward measurable goals using live product data.
- Why it matters
- Google pitches it at lean teams that struggle to balance post-launch maintenance, including security fixes, with building new features.
Google Labs announced Stitch Loop on 18 November, describing it as an experiment and part of a new direction for its Stitch design tool. Google says the agent runs over long periods, linking to a team's codebase and live metrics to spot problems and keep working until set goals are reached.
Teams connect GitHub plus product context such as logs, user behaviour and feedback, via more than 180 integrations, their own MCP servers or CLI uploads from a local machine. Stitch Loop then runs repeating loops. Each loop suggests opportunities backed by product evidence, offers one or more solutions with tradeoffs, builds the change in a separate cloud environment and submits it to GitHub as a pull request, then monitors live data to judge results. Google says nothing merges until a person reviews the pull request.
Three loops ship by default, covering cybersecurity, latency and accessibility, and users can add custom ones, such as improving checkout conversion. The security loop combines Gemini 4 Carbon with CodeMender, Google DeepMind's code-security agent, to scan for vulnerabilities and propose fixes.
Google says other coding agents, including Google Antigravity, can draw on Stitch Loop's product context through the Stitch CLI and MCP, and teams can use it in Slack. A waitlist is open globally, and eligible Google AI Ultra subscribers are first in line for access.