Google Stitch Loop: a product agent that keeps working after launch
Google Labs' Stitch Loop is a product agent that reads your code and live metrics, then ships fixes as GitHub PRs. Waitlist only, no price. Give it a metric.
Launch day is easy. Month six is where products rot. On November 18, Google Labs introduced Stitch Loop, a long-running product agent. It connects to your codebase and live product data, finds what needs fixing, and keeps working toward a goal you set. Google pitches it at lean teams: patch the vulnerability, speed up checkout, find why signups drop.
What actually happened
From Google's announcement:
- What it connects to. GitHub plus "logs, user behavior, feedback," through what Google says are over 180 integrations, custom MCP servers or CLI uploads. Google doesn't name the analytics tools.
- Three default loops. Security, latency and accessibility. You can add your own; Google's example is checkout conversion.
- How a loop runs. It suggests a change with evidence, offers options with tradeoffs, writes the code in an isolated cloud sandbox, and opens a GitHub pull request. Then it watches the live metric, and keeps watching after the goal is met in case it slips.
- The human gate. Google's words: "Nothing merges until you review the PR."
- Security loop. Pairs Gemini 4 Carbon with CodeMender, Google DeepMind's code-security agent.
- Slack. Mention @stitch for daily digests and to start loops.
- Availability. It's an experiment. The waitlist is open globally, with priority for eligible Google AI Ultra users. No price, no launch date, and nothing published on data handling.
Why it matters for your business
Small teams ship a site and move on. Nobody owns page speed or the form that broke on Safari. An agent that watches one number and sends PRs for review is a real fit for that gap.
The weak spot is the goal. "Improve the store" is a wish. "Cut mobile checkout abandonment from 70% to 60%" is a metric the agent can test against. Write the second kind.
Before you join the waitlist:
- Get your analytics right. The agent learns from the data you connect. Broken events give you confident bad PRs.
- Pick one metric and know its baseline today.
- Decide who reviews PRs and how fast. An agent that ships PRs 24/7 and a reviewer who checks weekly is a queue, not a loop.
- Ask about data before connecting logs with customer info. Google hasn't said how it handles them yet.
Key takeaways
- Stitch Loop launched as a Google Labs experiment on November 18
- Default loops: security, latency and accessibility; custom loops allowed
- Every change arrives as a GitHub PR; nothing merges without review
- Waitlist only, priority for AI Ultra; no price or data-handling terms yet
- Give it one metric with a baseline, not a vague goal
Know what a better checkout is worth before you automate it. Plug your traffic and conversion numbers into our ROI calculator, then talk to us about instrumenting the metric an agent can actually move.
Sources: Google: Introducing Stitch Loop, Stitch Loop waitlist.
- #google-stitch-loop
- #google-labs
- #ai-agents
- #github
- #conversion-optimization
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
OpenRouter drops 3 providers, pulls ZDR from 125 endpoints
OpenRouter off-boarded three inference providers and removed 125 endpoints from zero data retention after Dragos canary research. What to check in your stack.
Read itOpenAI Performance Factory: Codex agents debug, engineers merge
OpenAI's Performance Factory ran 1,700+ Codex investigations into ChatGPT slowdowns. 70% of audited root causes held up; about 15 agent PRs shipped as written.
Read it