Dynamic workflows in Copilot CLI and the Copilot app
GitHub Copilot now supports dynamic workflows in its CLI, app, and SDK, allowing developers to define multi-agent orchestration logic in code. These programs combine automated steps with agent tasks, supporting parallel execution, structured data passing, and human review checkpoints. This feature enables reusable, observable processes for complex tasks like incident investigation or parallel pull request reviews, distinguishing it from simple
Dynamic workflows are now offered in the Copilot CLI, the Copeteer app, and the Copilot SDK, letting developers script orchestration logic that blends automated steps with one or more agents. These programs, stored in a Copilot extension, let steps run sequentially, concurrently, or both, and they can invoke commands, tools, or external services, pass structured results between stages, have sub‑agents verify each other, request user input where supported, and pause at checkpoints for review before continuation. Typical uses include reusable processes such as incident investigation, parallel pull‑request file reviews, coordinated release checks with human validation, large‑scale codebase scans, and long‑running jobs that may need to be halted and resumed. Users may write workflows themselves or have Copilot generate them, with built‑in authoring guidance available. The feature is in public preview, works on all plans, and can be enabled in the CLI via the --experimental flag or the /experimental command; the app provides it without setup.