Google Stitch Loop Agent Ships 3 Loops and 180+ Integrations

Google Labs has introduced Stitch Loop, which it calls “a long-running product agent that connects to your codebase and live metrics to continuously improve what you’ve shipped”, in a blog post on 18 November. It is an experiment, and access starts with a waitlist.

The numbers in Google’s post: Stitch Loop connects to product context through “over 180 integrations”, ships with three default loops (security, latency and accessibility), and runs each loop through four steps, by our count of the steps the post lists.

Business Pill · Free Cash Flow

A Business Pill on free cash flow: the difference between profit on paper and cash in the bank. It is a general explainer and does not discuss Google.

The key insight: Google Labs’ Stitch Loop is a waitlist-only product agent that connects through over 180 integrations, ships three default loops and needs human approval for every pull request, per Google’s post.

What Google Announced

Google’s pitch is about the work after a feature ships. “Coding agents let you build and ship new features fast, but their job ends the moment your pull request merges”, writes Justin Toupin, Director, Product Management, Google Labs. Stitch Loop, the post says, “finds what needs fixing, and keeps working 24/7 on your product until your goals are met”.

Users connect GitHub and product context such as logs, user behaviour and feedback, “with over 180 integrations, custom MCP servers, or local CLI uploads”. The Stitch account on X describes it as “a product agent built for outcomes, not just PRs”.

Stitch Loop ships with three default loops and runs four steps per loop, per Google Labs' post; the step count
Stitch Loop ships with three default loops and runs four steps per loop, per Google Labs’ post; the step count is ours. Google also cites over 180 integrations.

What Was Posted on X

Google posted on 18 November 2026 that Stitch Loop, a new Google Labs experiment, keeps working after launch: using Gemini 4 Carbon and CodeMender, it watches live metrics, tests fixes in a sandbox and opens pull requests for users to approve, with a waitlist open.

How a Loop Works

Each loop, the post says, runs through four steps: Suggestions grounded in product evidence, Solutions with tradeoffs and scoping, Build, in which Stitch Loop “writes the code in an isolated cloud sandbox and opens a GitHub pull request for your approval”, and Reflect, in which it monitors live data and keeps watching after a goal is met.

Humans keep the merge. “You always have the final call. Nothing merges until you review the PR”, the post says. Teams can also create custom loops, such as one for improving checkout conversion.

The Security Loop

The built-in security loop “pairs Gemini 4 Carbon’s cyber security capabilities with CodeMender, Google DeepMind’s autonomous agent that improves code security”. That lands in a week when AI models are finding software flaws at speed: see our report on the nginx advisories and Bitdefender’s finding on poisoned GitHub repos.

Access and Workflow

The waitlist “opens today for all users, globally, with eligible Google AI Ultra users getting priority access”. Google says Stitch Loop can work as a service layer for coding agents such as Google Antigravity, through the Stitch CLI and MCP, and in Slack through @stitch. For another Google build tool this week, see AI Studio Build’s domain sales.

TestingCatalog, which says on X it has been testing the product for weeks, says its tests ran on Gemini 3.8 Flash; Google’s post names Gemini 4 Carbon only for the security loop.

The Structural Read

As we read Google’s post, the product moves the agent’s job from writing a feature to owning a metric after launch.

As we read it, the security loop ties Google DeepMind’s CodeMender to a commercial product surface for the first time in this post.

Justin Toupin, Google Labs

“Coding agents let you build and ship new features fast, but their job ends the moment your pull request merges”

Three Implications

FOR PRODUCT TEAMS Google says Stitch Loop opens pull requests and that nothing merges until a person reviews them.

FOR SECURITY The built-in security loop pairs Gemini 4 Carbon with Google DeepMind’s CodeMender, per the post.

FOR CODING-AGENT VENDORS Google positions Stitch Loop as a service layer other coding agents can pull context from through its CLI and MCP.

What Is Not Established

Google’s post gives no pricing, no launch date beyond the waitlist and no customer results. The list of 180-plus integrations is Google’s figure; we did not test the product. We did not contact Google.

This Week In The AI Supercycle

AI Lost Its Central Banker, the Lab Oligopoly, and the Future of Open Source

This week’s issue: the Nvidia leadership change, read as the AI build-out losing its central banker: who now finances the GPU landlords, backstops the data centres and sponsors open models, and why two labs are becoming the bank.

Read the full issue →

Business Engineer Framework

AI Unicorn Market Map

A Business Engineer map of where the largest private AI companies sit and how they raise.

Read the Map of AI →

The Bottom Line

Google Labs is extending Stitch from design into running products: a waitlist-only agent with 180-plus integrations, three default loops and human approval on every pull request.

97,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

A note on sourcing. Quotes and figures are from Google Labs’ 18 November blog post by Justin Toupin and Stitch’s post on X. The ‘four steps’ count is ours from the steps the post lists. TestingCatalog’s testing claim is attributed to it. We did not test Stitch Loop or contact Google.

Sources: Google Labs, Introducing Stitch Loop, 18 Nov 2026 · @stitchbygoogle on X, 18 Nov 2026 · TestingCatalog on X, 18 Nov 2026

Scroll to Top
FourWeekMBA