- GitHub says it saw more than 1.2 billion commits in one week this October, over 400% more than a year earlier. It cites a Faros AI study finding pull requests spent 441% longer in review when teams went from low to high AI adoption.
- The guide splits the workflow into four stages (Plan, Build, Verify, Observe) and lists Copilot features for each, including Plan Mode, cloud sandboxes, stacked pull requests, Agent Merge and Dreaming.
- The source gives no pricing or plan availability for these features, and GitHub package firewall is described as in public preview for npm.
What happened
On November 6, 2026, the GitHub Blog published a developer's guide to agentic engineering by Nick Liffen and Philip Holleran. It opens with a number: GitHub saw more than 1.2 billion commits in one week this October, which is over 400% more than a year earlier and more than GitHub received in all of 2025.
The argument is that faster code generation moves the pressure elsewhere. The post cites a Faros AI study saying pull requests spent 441% longer in review when teams went from low to high AI adoption. GitHub's answer is to use Copilot agents across four stages: Plan, Build, Verify and Observe.
Named features include Plan Mode, GitHub Agentic Workflows for labeling issues and estimating complexity, session continuity across devices, Copilot cloud sandboxes, Copilot Fusion for model selection, stacked pull requests, GitHub Code Quality, Copilot code review, Agentic Autofix, Agent Merge, and Dreaming for keeping memories and instructions current. The post does not say which plans include which feature.
What this means for founders
- Review is the bottleneck, not typing. If you are a small team shipping AI-written code, the post's point is that your pull request queue will grow. Smaller, focused PRs and automated checks are where it says to spend effort.
- Low-complexity work can go straight to an agent. The guide says to use Agentic Workflows to label issues and estimate complexity, then assign low-complexity tasks directly to Copilot. That could free a solo founder's time for harder decisions.
- Work can continue when you step away. Copilot runs tasks in cloud sandboxes and sessions can be picked up across clients, so a task does not have to stop when you close your laptop.
- Auto-merge is opt-in. Per the source, Copilot will merge pull requests once your verification steps pass only if you configure it. Decide what your checks need to cover before you turn that on.
- Costs and plan limits are unknown. The source does not give pricing, usage limits or which tiers get these features. Check before you plan around any of it. It also notes GitHub package firewall is in public preview for npm only.
What the Launch Llama directory shows
Live numbers from Launch Llama for listings that describe themselves as "ai code review", "ai coding assistant", "agentic coding", "code generation" tools, as of 2026-11-07.
| Signal | Launch Llama data |
|---|---|
| Listings in this space | 13 |
| New in the last 30 days | 1 |
| With a free or freemium plan | 70% |
| Most upvoted by founders | Claude Opus 4.7 (506 upvotes) |
Launch Llama lists 13 matching listings in this space, only 1 of them new in the last 30 days, and 70% have a free or freemium plan, so founders can test alternatives without paying up front.
How to try it today
- Pick one piece of work already on your plate, as the guide suggests, and start with the part that frustrates you most this week.
- For planning, use Plan Mode in your preferred Copilot surface to compare approaches and refine a plan. Review and approve it before handing it to Copilot.
- Ask Copilot to turn the plan into GitHub Issues, for example with this prompt:
Create GitHub Issues from this plan. Give each issue the context and acceptance criteria needed for implementation, and split out tasks that can be handled separately. - Break big changes into stacked pull requests so each piece can be reviewed on its own.
- Turn on Copilot code review and set the analysis depth yourself, or let Copilot choose per pull request.
- Once your checks are solid, consider configuring Copilot to merge pull requests when verification steps pass. Check availability for your plan first, since the source does not say.
Related tools on Launch Llama
Tools from the Launch Llama directory that fit with this news:
| Tool | What it does | Why it matters here |
|---|---|---|
| Self-Hosted Coding Agents for Kubernetes | For platform engineers: Self-Hosted Coding Agents runs autonomous agents in your Kubernetes cluster so you automate development work without vendor lock-in. | If you want coding agents without a hosted vendor setup, this runs autonomous agents in your own Kubernetes cluster, a different route from Copilot's cloud sandboxes. |
| Agent Callout | For developers: Agent Callout adds voice alerts to AI coding agents so you monitor progress without leaving your terminal. | The guide stresses letting agents keep working while you are away, and this adds voice alerts to AI coding agents so you can track progress from your terminal. |
| Agent Island | For developers: Agent Island monitors Claude, Codex, Grok, and Cursor sessions locally so you track AI coding progress without telemetry. | It monitors Claude, Codex, Grok and Cursor sessions locally, which may help if you run several coding agents alongside Copilot. The listing does not mention Copilot. |
Browse all tools or list your own tool.
Llama's take
I would try this now if you already pay for GitHub Copilot and your pain is a slow review queue or scattered context. Start with the cheap, reversible steps: Plan Mode, issues with acceptance criteria, and smaller pull requests. None of those require you to change how you work overnight.
I would wait if you need firm numbers. This is a GitHub guide, so it is a vendor's own framing, and the 441% review figure comes from a Faros AI study the post cites without detail. The source also gives no pricing or plan availability for the newer features.
The one thing to watch is auto-merge and Agent Merge. Letting an agent fix failing checks and merge is only as safe as your checks, so tighten those first.
FAQ
What is agentic engineering according to GitHub?
GitHub's guide describes it as using agents across the whole software workflow, not only code generation. It groups the work into Plan, Build, Verify and Observe, with Copilot features for each stage.
Can a GitHub Copilot agent merge pull requests on its own?
The source says that if you configure it, Copilot will merge your pull requests once your verification steps are passing. Agent Merge also addresses failing checks, review comments and merge conflicts to prepare pull requests for merge.
Does the guide say how much these Copilot features cost?
No. The source does not give pricing or say which plans include each feature. It only notes that GitHub package firewall is currently in public preview for npm.
