As reported on November 18, 2026, GitHub Agentic Workflows had become generally available. The open-source project runs coding agents in GitHub Actions for repository tasks such as issue triage and analysis of continuous integration failures. It entered public preview in June 2026.
Developers describe a task in natural-language Markdown and configure it with YAML front matter. The project compiles that definition into a standard Actions workflow, allowing teams to use existing runner groups and policy restrictions. Teams can use Copilot or connect other coding agents.
Write access is separated from the agent by default. An agent starts with read-only access; separate jobs, each with its own write token, apply proposed changes. GitHub says agents run in sandbox containers protected by Agent Workflow Firewall, and a threat-detection job scans proposed changes before they are applied. The project’s FAQ cautions that these safeguards do not automatically cover directly granted write access or custom jobs added by workflow authors. Authors can also require reviewer approval before a write job runs by adding GitHub Environment protection rules.
Since the public preview, GitHub has added support for Actions Runner Controller (ARC), its Kubernetes operator for self-hosted runners. ARC can use Docker-in-Docker to let an agent start sandbox containers inside a runner pod. The project guide notes a trade-off: ARC runs its Docker-in-Docker sidecar as a privileged container by default, placing a privileged component in the same environment as the least-privilege agent.
GitHub’s documented default spending cap is 1,000 AI Credits (AIC), or $10, per run, at $0.01 per credit. Its enterprise controls reference also lists defaults of 5,000 AIC per day, 400 AIC per threat-detection run and a 20-minute maximum for an agent step. Organizations can override those settings. The gh aw forecast command, no longer experimental, estimates monthly AI Credit spending, including threat-detection costs.
An Accenture senior manager, Ben Alton, said in GitHub’s announcement that one of the company’s leaders estimated a saving of 29 hours per week on development work alone. The announcement did not explain how that figure was measured. In a hands-on account of issue triage, freenet middleware team lead Sascha Mirow-Stanitzok took a more cautious approach: he considered the default budget too generous for that task and set limits of 100 AIC, 15 interactions and 10 minutes per run. He also warned that telling an agent to ignore instructions in issue text does not, by itself, protect a workflow; he would not use the same pattern for workflows that change code or create pull requests.
GitHub’s FAQ says conventional Actions remains the better choice for reproducible builds and tests, while agentic workflows are suited to tasks requiring content understanding or generation. The project is available to GitHub Free, Team and Enterprise users on github.com. Its default Copilot CLI engine requires a Copilot license.


