Iterate on tasks with Copilot CLI autopilot and fleet
The author used autopilot to iterate until done and /fleet to split work across parallel subagents, noting overhead on small tasks.
Done withGitHub Copilot
- What they did
- Daniel Ward turned on /experimental, then used autopilot mode, which loops until a large task is done without the ask_user tool. He pairs it with /plan for a guiding plan and /fleet, which spins up parallel subagents, including custom agents. He caps loops with a flag when he can't watch the run.
- How it went
- Autopilot lets Copilot finish a big task without repeated prompting, which suits quick proofs-of-concept. Fleet adds orchestration overhead that slows prompts, so small tasks may do better without it.
- Worth knowing
- Set --max-autopilot-continues if you can't monitor the run. The CLI recommends yolo mode for autopilot, and Ward avoids autopilot for production work because it makes review harder.
Try it yourself with GitHub Copilot
Using Copilot CLI in [repo path], work on [task] in autopilot mode and keep iterating until [definition of done, e.g. all tests pass]. If the task has clearly separate parts, such as [part A] and [part B], use /fleet to split them across parallel subagents; otherwise stay with a single agent. When you finish, summarize what was done and flag anything you weren't sure about.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: daninacan.com · Undated
Related use cases
Run a solo company with eight agent departments
A founder set up CEO, CFO, COO, lawyer, accountant, marketing, CTO, and improver agents that share memory and consult each other.
Assign website tweaks to Copilot coding agent
An author gave the Copilot coding agent two small changes on their own site; it opened draft PRs and the spacing change was good enough to merge.
Have Copilot coding agent write and push code
A developer found the agent good at context and code, but it pushed to a branch leaving main empty and spent hours checking after being told to proceed.
Build a C compiler with a team of parallel agents
Anthropic's engineering write-up on tasking Opus 4.6 agent teams with building a C compiler.
Troubleshoot a production-only performance issue
A developer used Cursor and Warp agents to analyze NestJS database queries and recommend indexes for a production performance problem.
Let terminal agents make phone calls
A CLI tool built on Twilio gives terminal agents a voice to make calls, with a dashboard for questions they can't answer.
Custom agent
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.