🤖GitHub Makes Agentic Workflows GA With Default $10 Per-Run Spending Cap
Your CI agents now have a spending limit
TL;DR
GitHub made Agentic Workflows generally available, letting coding agents run triage and CI failure analysis inside Actions. Agents start read-only, and write jobs need separate tokens. Default spending caps of $10 per run mean your bill won't explode overnight.
GitHub just made Agentic Workflows generally available, and the design is more cautious than the hype suggests. Agents run inside GitHub Actions for tasks like issue triage and CI failure analysis, but they start with read-only permissions. Any changes they propose get applied by separate jobs holding their own write tokens. If you run repo automation, that split is the thing to understand. If you're the person who gets paged when CI goes red, this changes your Monday. You describe the job in Markdown with YAML frontmatter, and the project compiles it into a standard Actions workflow that reuses your existing runner groups and policy constraints. Copilot works out of the box, and you can bring another coding agent. The catch: the protections don't automatically cover direct write permissions or custom jobs you add yourself. Put a GitHub Environment rule with a required reviewer in front of any write job. The money side is capped by default. AI Credits cost $0.01 each, and each run tops out at 1,000 AIC, or $10. Enterprise defaults add 5,000 AIC per day, 400 AIC per threat-detection run, and a 20-minute agent timeout. Run the gh aw forecast command before you turn anything on, since it estimates monthly spend including threat detection. Some engineers think the triage default is too generous, so you can drop it to 100 AIC, 15 turns and 10 minutes.

Key Points
Agents run in a sandboxed container behind the Agent Workflow Firewall, and a threat-detection job scans proposed changes before any of them are applied to your repository.
Keep reading — it's free
Enter your email to keep reading — plus the best of AI & tech, daily. Free, forever.
Already a member? Sign in
Comments
Be the first to comment