Plan then implement code changes with Copilot agent mode
A workflow where Copilot drafts a plan, and once approved via 'Start Implementation', Agent mode makes the changes automatically. ·
Done withGitHub Copilot
- What they did
- One commenter’s workflow was to use Copilot’s Plan mode for bigger changes so it first asked clarifying questions, then wrote a plan.md file for review. After the person approved it by clicking Start Implementation, Copilot switched into Agent mode and executed the plan by making code changes automatically; they also advised being specific and pointing it at files with #file.
- How it went
- The thread describes this as an effective way to handle large features while conserving paid requests, but it gives no concrete project example or measured result. Other commenters warned that request limits and context compaction could still interrupt runs.
- Worth knowing
- The clearest catch was cost control: several commenters said to treat premium requests as scarce and avoid wasting them by planning first and scoping context tightly.
Try it yourself with GitHub Copilot
I want to [describe change, e.g. add pagination to the orders list in [repo or file]]. First write a step-by-step plan listing files you'll touch and how you'll test it, and wait for my approval. After I approve, switch to agent mode, make the changes, run [test command], and summarize what changed.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: Reddit · Apr 1, 2026
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.
Migrate a large web application with Copilot Agents
A developer found Copilot Chat struggled with a large web app migration, so switched to Copilot Agents, which handled the codebase better; the migration was still early.
Run terminal commands via Copilot CLI subagents
A commenter uses Copilot CLI agents that execute terminal commands, including the /fleet command and subagent-style workflows.
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.
Automate web configuration from coding sessions
Commenters use Claude in Chrome to carry out web configuration requests that arise from coding sessions and other browser operations.
Turn GitHub issues into draft pull requests with long-running agents
The author used Dots for read-heavy long-horizon tasks like reading GitHub issues and making PRs, which were typically poor but gave a starting point.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.