Run longer coding sessions with Copilot Agent Mode
A developer moved from Copilot syntax questions to agent-mode sessions, switched to Claude Sonnet as the model, and tried building more complete things.
Done withGitHub CopilotClaude
- What they did
- After a GitHub tutorial attempt stalled, Jonas started a repo for a web app to track scores for his Partners board game group. He added an outline and his existing JSON data, then used VS Code agent mode with Claude Sonnet. The agent built score entry, a history table, statistics charts, avatar mood pop-ups, Danish/English support and dark mode. He made only a few manual edits.
- How it went
- He got a working app, deployed on Digital Ocean App Platform. The tutorial attempt ended in a loop of the agent undoing its own changes. The hover pop-up caused similar trouble, and the repo needs cleanup of leftover agent files.
- Worth knowing
- The agent often lost its link to the editor and reported its own port as in use, which cost time. Commit after each working step, which he forgot to do.
Try it yourself with GitHub Copilot
Use agent mode with Claude Sonnet as the model to build [a small but complete feature or app, e.g. a CRUD page for X] in my project in [folder or repo]. Work in stages: first propose a plan and file list, then implement it, run the app or tests, and fix any errors you hit. Finish when the feature works end to end, and give me a summary of the files you changed and how to run it.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: dev.to · Undated
Related use cases
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.
Fix Shopify SEO and find ad waste with a Claude agent
A Claude-based agent analyzed a Shopify store, fixed SEO issues raising the score to 95, found ad waste, tested ideas and checked AI visibility.
Coach training and diet from wearable data
The poster uses Claude with Apple Health, nutrition and lifting app data, plus medication and bloodwork files, and has it maintain an HTML dashboard.
Turn tickets into merged pull requests
Pulls tasks from GitHub Issues or Linear, then runs Claude Code or Codex in isolated Kubernetes pods to carry each one through to a PR.
Turn customer feedback into tested pull requests
In OpenAI's own launch example, a Dot watches feedback for recurring requests, scopes small fixes, builds and tests them, and returns PRs with demo videos.
Code in the terminal with an agent
The poster uses Grok as a terminal coding agent that reads, edits, runs tools and spawns sub-agents.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.