GitHub Stacked PRs Hit Public Preview: Small Reviews for the Agent Era

TL;DR
GitHub stacked pull requests are in public preview: how stacks split one large change into small, reviewable PRs with one-click merge and a gh-stack skill.
Last updated: July 31, 2026
What shipped#
On July 30, GitHub put stacked pull requests into public preview for all repositories. A stack is an ordered series of pull requests where each PR targets the branch of the PR below it, forming a chain that ultimately lands on your default branch. It ships as a GitHub-native workflow plus a CLI extension:
gh extension install github/gh-stackcreates and manages stacks from the terminal- Stacks also work on github.com, the GitHub mobile app, and with coding agents that use the gh-stack skill
- Each PR shows a stack map at the top, so reviewers see how their layer fits into the larger change
- Any layer can be reviewed in isolation, showing only that layer's diff
- Merge one, some, or all: landing the bottom-most ready PR brings every merged layer below it, and open layers above stay open and auto-rebase and retarget
- Existing branch protections and required checks still govern what reaches
main
The preview is rolling out to all repositories "over the coming days," and merge queue support is rolling out progressively over the coming weeks, per the changelog.
The companion feature is stacked sessions in the GitHub Copilot app. An agent starts a new session on top of an existing one: it takes the previous context, branches off the existing session's work, plans, and opens its own pull request targeting the first PR's branch. The GitHub blog walkthrough shows the pattern end to end: a decade-old React 15 codebase gets a styling modernization PR, then a react-bootstrap removal PR stacked on top, each independently reviewable, landing as one unit.
Why it matters#
This is a direct answer to the bottleneck that agent-era development created. Coding agents generate large diffs faster than teams can review them - the exact problem we covered in AI Code Review Is the New Bottleneck and its follow-up, AI Coding Agents Move the Bottleneck to Review Queues. Stacks are the mechanical fix: narrow, dependency-ordered pieces that reviewers can check in parallel without blocking later work.
The changelog's quotes from named maintainers underline that this is a real production pattern, not a demo. TED CTO Andy Merryman: AI made his developers more productive, "but that created a new bottleneck: PRs were growing large enough that reviewers were struggling." Next.js lead Tim Neutkens says stacks helped his team at Vercel "introduce smaller individual changes while shipping larger features, making it easier to review PRs." jQuery creator John Resig reports landing "5 stacked PRs directly to a merge queue all at once."
My take#
Three things make this more than a nicety.
Native beats bolted-on. Until now, stacking meant third-party tooling or hand-rolled rebase choreography. Because stacks are built into GitHub, every existing review, check, and merge requirement works out of the box, which lowers the cost of switching to near zero. The default path to a small, reviewable PR just got shorter.
Agent-native by design. The gh-stack skill is the notable piece: a coding agent can open a stack without generating dozens of rebase commands. This is GitHub shipping the workflow that agents already need, and it fits the pattern we documented in Agent PR Governance: The New Rules for Copilot Reviews: the guardrails that kept agent PRs sane are now applied to the granular version of them.
Merging is where it pays off. One-click landing for the whole chain, with merge queue support coming, removes the choreography tax that made manual stacking unattractive. A stack of five small PRs that lands like one merge is a genuinely new primitive, closer to how git forges are being rebuilt for agents than to an incremental UI tweak.
The caveats are real. Public preview means behavior can change. Per-layer required checks multiply CI runs. Stacks assume linear, ordered work, so a hotfix that lands in the middle still needs the normal dance. And the auto-rebase path only helps if layers stay small - a stack of five 2,000-line diffs is just a large PR with extra steps.
Still, the direction is clear. The same logic that makes small commits cheap for a parallel fleet - we run one here - applies to review. Stacks give teams a first-class way to keep changes small without losing the atomic land, and the merged-stack workflow that 12 Tools in One Night showed agents already want is now the vendor default.
FAQ#
What are GitHub stacked pull requests?#
A stack is an ordered series of pull requests where each PR targets the branch of the PR below it, forming a chain that ultimately lands on the default branch. GitHub put stacked PRs into public preview for all repositories on July 30.
How do I create a stack from the terminal?#
Install the CLI extension with gh extension install github/gh-stack. Stacks also work on github.com, the GitHub mobile app, and with coding agents that use the gh-stack skill.
What happens when I merge part of a stack?#
You can merge one, some, or all layers. Landing the bottom-most ready PR brings every merged layer below it, while open layers above stay open and auto-rebase and retarget. Existing branch protections and required checks still govern what reaches main, and merge queue support is rolling out progressively.
How do stacked PRs help with AI coding agents?#
Coding agents generate large diffs faster than teams can review them, and stacks split one large change into narrow, dependency-ordered PRs that reviewers can check in isolation. The gh-stack skill lets an agent open a stack without generating rebase commands, and stacked sessions in the GitHub Copilot app let an agent branch off an existing session and open a PR targeting the previous one.
Continue Reading#
- Agent PR Governance: The New Rules for Copilot Reviews
- AI Code Review Is the New Bottleneck
- AI Coding Agents Move the Bottleneck to Review Queues
- Cursor Origin: A Git Forge Built for AI Agents, Not Humans
- 12 Tools in One Night: An Honest Overnight Agent Report
- PAIChecker: 13.6% of SWE-bench Verified Instances Have Misaligned PR-Issue Pairs
- Coding Agents Tripled PRs. Total Dev Time Still Went Up - Linear's usage data: agent-connected teams went from 21 to 65 weekly PRs while total product-development time still rose
Sources#
Get the next deep dive like this in your inbox
One email a week on News and the rest of the AI dev stack. Free.
Read next on AI coding tools
Agent PR Governance: The New Rules for Copilot Reviews
Agent PR governance for Copilot code review: a five-layer policy stack for agent-authored pull requests, from agent validation and review depth to attribution.
8 min readAI Code Review Is the New Bottleneck
AI code review is the new bottleneck as coding agents outpace reviewers. Build review systems around reproduction, small diffs, tests and receipts.
8 min readAI Coding Agents Move the Bottleneck to Review Queues
AI coding agents move the bottleneck from code generation to review queues, CI minutes, environment reliability and merge discipline. What teams must fix next.
8 min readNew here? Start with
Technical content at the intersection of AI and development. Building with AI agents, Claude Code, and modern dev tools - then showing you exactly how it works.







