Browse the guide
Demos
Agents Page
Start, monitor, and steer Copilot coding agents across repositories from the Agents page.
What This Demonstrates
The Agents page (github.com/copilot/agents) is GitHub’s control plane for the Copilot coding agent — start agent tasks across repositories, watch their sessions live, steer them mid-run, and review the pull requests they open. This demo uses Copilot to plan how you’d put that orchestration to work on this repo.
Prerequisites
- GitHub web UI (the Agents page at
github.com/copilot/agents) - Repository with existing PR history for reference
Prompts
Prompt A — PR Automation
For this repo, propose 3 specialized agents I should configure to run on thisrepo: one for Terraform/Azure, one for backend API quality, one for frontend UXconsistency. For each, define scope, trigger conditions, and expected outputs.What to look for:
- Agent scopes matching the directory structure
- Trigger conditions (file path patterns, PR labels)
- Expected outputs (comments, status checks, suggestions)
Prompt B — Multi-Agent Orchestration
Simulate a multi-agent workflow for a PR that touches `infrastructure/terraform`and `services/order-service`. Show which agents should run, in what order, andwhat each should validate.What to look for:
- Agent ordering (infra validation before service checks)
- Each agent’s specific validation scope
- Conflict resolution between agent outputs
Prompt C — Policy Alignment
Draft branch protection + agent policy recommendations for this repo so thatinfra changes, security checks, and service contract changes are all enforcedbefore merge.What to look for:
- Required status checks tied to agent results
- Path-based rules (e.g.,
infrastructure/**requires infra agent approval) - Escalation for security-sensitive changes
Follow-Up Prompts
- “Create a CODEOWNERS file that aligns with these agent policies.”
- “What metrics should I track to measure agent effectiveness over time?”