Woman in a mustard sweater holds a glowing cube in both hands in a dark room, beside her a floating green ring of light with orbiting cubes and a brightly lit open door, symbolic image for Google's Stitch Loop product agent (AI image)

Illustration

New tool2 min read

Google Labs opens waitlist for Stitch Loop, a 24/7 product agent

Google Labs has unveiled Stitch Loop, a product agent that reads a product's code and live metrics, suggests improvements and opens GitHub pull requests for them; Google announced the experiment in a blog post on 18 November. The waitlist is open to all users worldwide, Google says, with priority for Google AI Ultra subscribers, and the company names no price. Nothing merges without sign-off: a human has to approve every change in the pull request.

Key facts

  1. Google Labs announced Stitch Loop on 18 November 2026, an agent that reads code and live metrics, makes suggestions and opens GitHub pull requests.
  2. Three loops ship by default, for cybersecurity, latency and accessibility; Google says nothing merges until a human has reviewed the pull request.
  3. The waitlist is open worldwide to all users, Google says, with priority for Google AI Ultra subscribers; Google gives no price.

Manfred „Luigi“ LugmayrBy Manfred „Luigi“ Lugmayr

Researched and written with AI, to Luigi's brief, positions and editorial rules. Automatically checked against the sources. How I work →

According to Google, Stitch Loop runs a repeating four-step cycle. It surfaces opportunities backed by evidence from the product and offers one or more solutions for each, with trade-offs and scope. Once the team picks a direction, the agent writes the code in an isolated cloud sandbox and opens a GitHub pull request, with screenshots or videos where relevant. It then monitors live product data, learns from positive and negative results, and keeps watching after a goal is met in case the metric slips. Nothing merges until the team has reviewed the pull request, Google says.

Stitch Loop needs a GitHub connection. Product context such as logs, user behaviour and feedback comes in through more than 180 integrations, custom MCP servers or local command-line uploads. Google ships three loops by default: cybersecurity, latency and accessibility. Custom loops are possible, and Google names improving checkout conversion as an example. The security loop, Google says, pairs the cyber capabilities of Gemini 4 Carbon (our report on Gemini 4 Carbon) with CodeMender, a Google DeepMind agent for code security. The loop flags each issue with evidence and opens a pull request with a proposed fix.

Coding agents and Slack

Stitch Loop can also act as a service layer for other coding agents, such as Google's Antigravity. Through the Stitch command-line tool and MCP they can pull the stored product context, Google writes. In Slack, the @stitch command delivers daily insight digests and creates new loops.

Much stays open. Google gives no price and no measurements or customer examples showing how good the suggestions are. The post names no country restriction for the waitlist. Google itself calls Stitch Loop an experiment. How far Google's agents may act without asking is also a question for Gemini computer use on the Mac.

Analysis

The real step forward is less the code writing than the closed circuit: an agent that keeps watching the numbers after launch is a different tool from an assistant whose work ends at the merge. That also marks the limit. The suggestions are only as good as the metrics the agent watches. A team that has not defined events, goals and tracking cleanly will get pull requests built on weak data. Weaknesses in data and processes that already exist are more likely to grow than shrink with an agent like this.

The second question is sign-off. A loop such as "improve checkout conversion" touches things marketing often owns: copy, forms, price display. The only check Google describes is the pull request. Whether a developer reviews it alone or marketing and product join in is something a team should settle before the first run. A narrow first case is a sensible start: one goal, one metric, one reviewer. For setting up agents in marketing, see our guide to AI agents for marketing; for how far agents already get on multi-hour tasks, see the OpenAI report on AI agents.

  • Share

Got AI news? Submit news →

FAQ

What is Google Stitch Loop?

Stitch Loop is a Google Labs experiment: an agent that reads a product's code and live metrics, suggests improvements and opens GitHub pull requests. Google announced it on 18 November 2026.

How much does Google Stitch Loop cost?

Google's blog post gives no price. Google says the waitlist is open to all users worldwide, with priority for Google AI Ultra subscribers.

Does Stitch Loop change code without approval?

Not according to Google: the agent opens a pull request, and nothing merges until a human has reviewed it.

Source

Google Labs: Introducing Stitch Loop: A new way to build and evolve products (18.11.2026)

Black Friday 2026

Black Friday deals on AI tools, straight to your inbox.

I collect the AI tool deals worth having and send them in the AI Briefing, along with the AI news that matters, every Wednesday and Friday. Five minutes to read.

  • Black Friday deals on AI tools, collected for you
  • The key AI news, briefly explained
  • Wednesday and Friday, unsubscribe in one click

In the AI Briefing right now

Image for the story: Superlist Black Friday: Super plan costs $99.50 in year oneDealSuperlist Black Friday: Super plan costs $99.50 in year one
Image for the story: Google Labs opens waitlist for Stitch Loop, a 24/7 product agentNewsGoogle Labs opens waitlist for Stitch Loop, a 24/7 product agent
Image for the story: Intuit to sell Mailchimp to Bending Spoons for $3.68bnNewsIntuit to sell Mailchimp to Bending Spoons for $3.68bn
Blonde woman with glasses smiling as she scrolls on her phone in an office

Manfred “Luigi” LugmayrFrom Manfred “Luigi” Lugmayr: 30 years in marketing, seven of them as CMO at burgerme.