GitHub Copilot enterprise-managed agent operation permissions settings Image: GitHub / github.blog
by Michael Joiner

GitHub Copilot Gives Enterprise Admins Control Over What Agents Can Do Without Asking

A new enterprise-only setting lets admins decide which Copilot agent operations are blocked, which need human approval, and which can run automatically — and users can't override it.

Share

GitHub Copilot rolled out a new enterprise admin control on September 9 that lets organizations set centralized rules for what agents can do without asking for permission.

The feature is called Enterprise Managed Permissions for Agent Operations. It’s available now for Copilot Business and Copilot Enterprise.

What It Controls

Admins can define three behavior levels for each category of agent action:

  • Block — the agent is not allowed to take this action at all
  • Require approval — the agent must pause and ask a human before proceeding
  • Allow — the agent can proceed automatically

The categories covered are shell commands, file reads and writes, and network domains. Each can be set independently.

The key enforcement detail: these restrictions cannot be overridden by user settings, workspace configurations, or previous approvals. If an admin blocks shell commands, no user in that organization can grant the agent permission to run them.

Admins can also create different policies for different teams within the enterprise, so a security team working with sensitive repositories can have tighter controls than an internal tools team.

The policy works across the GitHub Copilot app, Copilot CLI, and VS Code with Agent Host.

Why This Matters

Copilot has been adding more agentic capabilities throughout 2026 — agents that can read your codebase, execute terminal commands, push code, and open pull requests. That’s powerful for productivity, but it creates a real governance problem for companies that need audit trails and approval workflows before anything touches production.

This control is a direct answer to that. Enterprise IT teams can now decide exactly where the autonomy boundary sits without relying on individual developers to set their own limits.

GitHub’s changelog for the release is at github.blog/changelog.

Share