GitHub Copilot Introduces Dynamic Workflows for Multi-Step, Agent-Coordinated Tasks

GitHub launched dynamic workflows across Copilot CLI, the GitHub Copilot app, and the Copilot SDK — code-based programs that orchestrate multi-step tasks by combining automated operations with AI agent involvement.

Details

• What they are: a program that defines how a task is carried out, combining automated steps with the work of one or more agents, living inside a GitHub Copilot extension and built on Copilot’s extensibility APIs
• Capabilities: running commands, tools, or external services; executing independent tasks in parallel or sequentially; passing structured data between stages; incorporating agent verification and human input; and pausing at checkpoints for review before resuming
• Vs. /fleet: unlike /fleet, which delegates work to subagents dynamically, dynamic workflows execute a predetermined process structure
• Use cases: multi-stage release validation with human review gates, parallel analysis of pull request changes, pattern detection across large codebases, research-to-implementation sequences, and long-running operations needing pause/resume support
• Availability: public preview across all Copilot subscription tiers — enabled by default in the GitHub Copilot app, available via the --experimental flag or /experimental on in Copilot CLI, and included in the Copilot SDK

What happened next

Dynamic workflows land alongside GitHub’s new desktop “computer use” feature and continue Copilot’s push from single-shot code suggestions toward structured, auditable multi-step automation — complementing rather than replacing the more freeform multi-model orchestration GitHub has been previewing with HydraFusion.