Move from a checked suggestion to a scoped IDE agent change.
Coding & development · Documented · Documentation checked 2026-10-04
Editor assistance and agents, GitHub cloud agent, and code review have different execution contexts.
Plan, supported IDE, repository permission, AI credits, and organization policy determine access.
An illustrative recipe. This website does not run the automation.
The goal and the right kind of system
Move from a checked suggestion to a scoped IDE agent change.
Use ordinary chat for explanation and Agent mode only for the multi-step implementation.
Inputs and connected services
Inputs: supported IDE, sample repo, reproduction, project instructions, and test criteria.
Services: GitHub Copilot in the selected IDE and local tools.
The steps
An illustrative sequence to adapt and test before using real data or enabling ongoing work.
- Inspect the function in chat and confirm expected behavior.
- Select eligible Agent mode for the scoped change.
- Review streamed edits and command approvals.
- Run checks, read the diff, and preserve human commit review.
Outputs and human checkpoints
Outputs: fix, actual verification, and clear untested limits.
Human checkpoints: compare suggestions with code and inspect the final change.
Exceptions, permissions, and stopping conditions
Permission boundary: sample workspace; no tracker or repository-provider writes.
Stop on missing Agent access, unavailable dependencies, failed checks, or unrelated edits.
Adaptable prompt
Fix [sample reproduction] in this IDE workspace. Follow project instructions, preserve unrelated edits, and report actual tests. Review necessary commands. No push, merge, deployment, or connected-service writes.
Related reading
Completions, chat, and editor context
Distinguish a code suggestion from an agent that can edit files and run commands.
IDE agent mode and tool approvals
Use an interactive agent for a bounded multi-file change and review each consequential command.
Instructions, custom agents, and MCP
Combine repository guidance with explicitly authorized tools, without treating instructions as security controls.
Designing an AI workflow
Make a useful task repeatable with clear inputs, steps, outputs, exceptions, and human checkpoints.
Permissions and meaningful approvals
Decide what a system may read, what it may change, and when someone needs to authorize a concrete action.
Sources and verification
Documentation checked 2026-10-04. Examples are illustrative unless described otherwise.
About GitHub Copilot · GitHub
Plans for GitHub Copilot · GitHub
Use agent mode in your IDE · GitHub
Repository custom instructions · GitHub
Enhance agent mode with MCP · GitHub