On 28 October 2026 GitHub Next announced that GitHub Agentic Workflows is generally available, after a public preview we covered in September. It lets teams automate reasoning-heavy chores such as issue triage, CI failure analysis, documentation updates and test writing with coding agents running inside GitHub Actions.
How it works
Automations are written as natural-language Markdown files and compiled into standard Actions YAML. They run on your existing runners and policies, powered by GitHub Copilot or your own coding agent. The project is open source.
The safeguards GitHub lists
Agents run read-only by default, inside a sandboxed container behind a firewall. Outputs pass a “safe outputs” check, and a separate threat-detection job scans proposed changes before they are applied. Since preview, GitHub added support for self-hosted runners, stronger isolation for cross-repository work, per-workflow model choice, model routing by task difficulty, and a gh aw forecast command that estimates AI credit spend.
Availability
It is generally available on github.com for Free, Team and Enterprise. The default engine needs a Copilot licence.
What it means
The unit of AI development is moving from a developer’s chat window to background automation on the repository. That is where the productivity gain is, and also where an unreviewed agent can do the most damage, which is why the safeguards are the story as much as the feature.
So what
Before agents touch your codebase, decide what they may read, write and merge, and who signs off. Our AI-assisted development service sets those boundaries from day one. If you want it applied to your product, start a project.