Google's Stitch Loop: an always-on agent that proposes security, speed and accessibility fixes
Google Labs has opened a waitlist for Stitch Loop, an experimental agent that watches your code and product data and opens GitHub pull requests for security, speed and accessibility fixes.
- Author
Umair Mateen- Published
- Reading time
- 3 min
- Category
- AI
Ideas in this post
Tap one for a 5-second answer. Underlined words in the article work the same way.
0 of 4 understood

Google Labs has opened a waitlist for Stitch Loop, announced on 18 November 2026. It is an experimental "product agent": an AI that stays connected to your code and your product data, looks for problems, and proposes fixes as code changes for your team to review.
What it actually does
Google describes Stitch Loop as "an always-on product agent from Google Labs". You connect it to your GitHub code, and it gathers context from logs, user behaviour, feedback and other connected tools. It then runs "loops": repeated checks that look for a kind of problem and suggest changes.
It ships with three default loops:
Quick check · Spot the weak point
Where is this agent's workroom weakest?
A company runs its AI coding agent in a sandbox set up like this. Which part is the weak point?
From the explainer What is a sandbox? Explained with a locked workroom for a contractor
- Security: finds and fixes vulnerabilities, combining Gemini 4 Carbon's cyber skills with Google's CodeMender tool.
- Speed (latency): looks for slow parts of the product.
- Accessibility: looks for problems that make the product harder to use for people with disabilities.
Teams can also set up their own loops. When it finds something, it writes the code in an isolated (a closed-off test environment) and opens a pull request on GitHub. Google says nothing merges until you review the pull request.
What it connects to
Besides GitHub, Google lists over 180 integrations, custom servers (a standard way to plug tools into ) and local file uploads through a command-line tool. It works with Google Antigravity and the Stitch command-line tool. In Slack, you can mention @stitch to get daily digests and start new loops.
Who can get it
The waitlist is open to everyone, worldwide, with priority for eligible Google AI Ultra subscribers. Google hasn't announced pricing or a date for wider access. This is a Labs experiment, so features may change or be dropped.
Who should care
Product and engineering leads at small teams that know they have a backlog of security, speed and accessibility fixes but no one assigned to work through them. It is also relevant to anyone weighing whether an AI agent should have standing access to their code and production data.
Limits and caveats
- It is waitlist-only and experimental, with no published price or service guarantees.
- Google gives no accuracy or results figures, so you'll need to judge the quality of its pull requests yourself.
- Giving an agent access to code, logs and user data is a security decision. Decide what it can see before connecting it.
- Accessibility fixes from an agent are a start, not a compliance check. If you have legal obligations, keep your usual audit.
What to do this week
- If you're curious, join the waitlist; it costs nothing to sign up.
- Decide in advance which repository and data sources you'd let it see, starting with something low-risk.
- Make sure branch protection on GitHub requires a human review before merge, for any AI agent, not just this one.
- Write down how you'll judge it: how many of its pull requests you'd actually accept.
The DevDrift helps teams set up AI agents in their development workflow with review steps and .
Sources
You’ve read the news. Now understand it.
- Learn · 5 min · Spot the weak pointWhat is a sandbox? Explained with a locked workroom for a contractorA sandbox is a closed-off space where a program or AI agent can work freely but can't touch anything outside it. Explained with a contractor given one locked workroom instead of a master key.Start in 5 seconds →
- Learn · 6 min · Myth or factWhat is MCP? Explained with an office of service countersMCP (Model Context Protocol) is one shared way for AI assistants to connect to other software and use it with your permission. Explained with an office where every department takes the same request slip.Start in 5 seconds →
- Learn · 5 min · Myth or factWhat is an AI agent? Explained with an assistant you send on errandsAn AI agent takes a goal, plans the steps and carries them out with software tools, checking each result as it goes. Explained with an assistant you send on errands.Start in 5 seconds →
- Learn · 5 min · Myth or factWhat is an audit trail? Explained with a visitor logbookAn audit trail is a time-ordered record of who did what in a system, when and with what result. Explained with the visitor logbook at an office reception desk.Start in 5 seconds →


