Stitch Loop: Google’s Always-On Agent for Your Live Product

Human Verified 0

Share to: , and

Developer desk at dusk with a glowing loop of arrows above a monitor showing a product dashboard and a small robot assistant

Google has introduced Stitch Loop, a long-running product agent that keeps working on software after it ships, watching live metrics and code and filing fixes for review. The Google Labs experiment opens its waitlist today, 18 November, to users worldwide, with Google AI Ultra subscribers getting priority access.

How Stitch Loop works

Stitch Loop connects to a team’s GitHub repository and product context, from logs and user behaviour to feedback, through more than 180 integrations, custom MCP servers or local command-line uploads. Grounded in that context, it runs evergreen loops aimed at measurable outcomes the team sets.

Each loop runs through four steps. First it surfaces opportunities grounded in product evidence, then proposes solutions with trade-offs and scoping, and visual options where they apply. It then writes the code in an isolated cloud sandbox and opens a GitHub pull request for approval, then keeps watching.

The security loop

Stitch Loop ships with three built-in loops for cybersecurity, latency and accessibility, and teams can add custom loops, such as improving checkout conversion. Security is the sharpest of the three, and Google has given it the strongest backing of any loop in the experiment.

The security loop pairs Gemini 4 Carbon‘s cyber security capabilities with CodeMender, Google DeepMind’s autonomous agent for improving code security. Together they scan a codebase for vulnerabilities before they become incidents, flagging each issue with evidence and opening a pull request with a proposed fix.

Works with your agents

Stitch Loop can also act as a service layer for coding agents such as Google Antigravity, or sit inside the places teams already work, such as Slack. Any coding agent can pull persistent product context from Loop through the Stitch command-line tool and MCP, powering mixed-model setups.

In Slack, teams can use @stitch for daily insight digests and to create new loops. Nothing merges without a human decision: every change arrives as a pull request, and Google stresses that the final call always stays with the team.

Our Read

Coding agents have a stopping problem: their job ends the moment a pull request merges, yet the real work of a product starts after launch, in patching, speed and drop-off diagnosis. Stitch Loop is Google’s answer to that gap, moving from generating interfaces to owning the maintenance loop itself.

The security pairing is the part to watch. Gemini 4 Carbon was built with cyber defence in mind, and CodeMender is DeepMind’s code-security specialist, so this loop has real capability behind it. The open question is quality: whether teams merge the fixes Loop files, or learn to ignore them.

Why It Matters

For lean teams, post-launch maintenance is the work that never ends, and it is also what separates products that improve from ones that quietly decay. An agent that watches live metrics and files fixes could let a small team maintain a product like a much larger one.

If the security loop works as advertised, vulnerabilities get fixed before they become incidents, not after. That is the promise; the waitlist means the proof will come from the teams who get in first, and from how many of Loop’s pull requests they actually accept.

Background

Stitch began as Google Labs’ tool for turning ideas into working interfaces in minutes, and Google says millions of designers, developers and founders now use it. Loop extends that story past the launch, into the long tail of keeping a product healthy.

Google has spent 2026 pushing agents into every part of the workflow, from Antigravity’s coding teams to Gemini Tasks, made free for hundreds of millions this week. Stitch Loop is the next step: an agent that does not just build the product, but stays to look after it.

Source: Google’s announcement

Share to: , and

Faheem R. Khattak

Senior automation engineer and AI marketing journo covering LLMs, algorithms, analytics, core updates, Bing search, and technical SEO.

0 Comments

Leave a Reply

Optional Take Tag:

🛡️ Spam protection: Honeypot + Time-Trap Active

This site uses Akismet to reduce spam. Learn how your comment data is processed.