Run multiple agents at once with /fleet in Copilot CLI

The GitHub Copilot CLI just got a powerful new capability: the /fleet command. Instead of running one agent at a time to solve a problem, you can now dispatch multiple agents in parallel to work on different parts of your task simultaneously. This is a meaningful shift in how you can interact with AI-assisted development—from sequential, single-threaded assistance to concurrent, distributed problem-solving. If you’ve ever wished you could have Copilot work on multiple files or components at the same time, /fleet is designed for exactly that scenario.

Here’s how it works technically. When you invoke /fleet, you’re telling Copilot CLI to analyze your prompt and intelligently split the work across multiple agent instances. The key is in how you write your prompts: you need to be explicit about what’s independent and can run in parallel. For example, instead of “refactor my API handlers,” you’d frame it as “refactor the auth handler, payment handler, and logging handler in parallel.” Copilot then creates separate agents, each focused on one piece, and coordinates them to work together. You can also declare dependencies—essentially saying “don’t start task B until task A is complete”—which prevents race conditions or logical conflicts. Behind the scenes, this is coordinated orchestration; Copilot needs to understand the scope boundaries and merge the results intelligently.

The practical value becomes clear when you think about real workflows. Imagine you’re updating an API with multiple endpoints that follow the same pattern. Instead of asking Copilot to handle them one at a time (and waiting for each to complete), you can ask /fleet to update your users endpoint, products endpoint, and orders endpoint concurrently. Or consider migrating a codebase from one authentication library to another across five different service modules—/fleet can tackle each module in parallel. For data processing pipelines, testing suites with independent test files, or infrastructure-as-code deployments with multiple components, parallel agent work can significantly reduce iteration time and keep momentum on larger refactoring tasks.

To avoid pitfalls, keep your prompt granular and clear about boundaries. Vague prompts that could be interpreted multiple ways will confuse parallel agents. Be specific about file paths, function names, and expected behavior. Also, remember that while agents work in parallel, they still need to be logically independent or properly sequenced—if handler B depends on a utility function that handler A creates, make that dependency explicit. Like distributed systems in general, the coordination logic matters as much as the individual agents. As this feature matures and you get comfortable with it, you’ll find it becomes an essential tool for accelerating large-scale refactoring and multi-component development work.