GitHub has made its assisted permissions feature generally available across Copilot CLI, the Copilot app, and Visual Studio Code, handing an artificial intelligence model the authority to automatically approve certain commands that an AI coding agent wants to run.
The change, announced in a GitHub changelog post dated October 26, 2026, shifts how developers manage the growing autonomy of AI coding agents operating inside their codebases.
How the new approval system works
When Copilot requests permission to run a command, access a file system path, or open a URL, a large language model evaluates the request. Requests classified as clearly safe automatically proceed. Other requests require your approval.
Assisted permissions sit between the two modes everyone already knows, approving every single action manually, or granting blanket permission that removes oversight of individual steps.
An independent analysis of the release described the tradeoff between speed and control.
Developers can turn the feature on differently depending on which surface they use. In Copilot CLI, run /permissions, then select assisted.
In the GitHub Copilot app, open the mode picker, click Tool permissions, then select Assisted. In Visual Studio Code, open the permissions picker, then select Assisted permissions.
Guardrails for enterprise customers
If you’re an enterprise administrator, you can use enterprise-managed settings to block Allow All while keeping Assisted available or disable Assisted separately.
Administrators can also choose the default permission mode for new sessions. Developers can still select another mode permitted by policy. If the default is unavailable or prohibited by policy, the session starts with manual approval.
For organizations running unattended workflows, the feature also changes how automation pipelines behave. According to GitHub’s documentation, it overrides –allow-all-tools when successfully enabled, so tool requests use assisted permissions instead of blanket tool approval.
If assisted permissions can’t be turned on, because enterprise settings block it for example, the CLI writes a warning to standard error and continues.
Why the timing matters
The general availability rollout arrives as GitHub pushes Copilot further toward autonomous, multi-step execution rather than single suggestions. The company’s terminal-native coding agent reached general availability in February 2026.
GitHub has since layered in features such as a tabbed terminal interface and deeper Model Context Protocol support.
As agents handle more of the software development lifecycle unsupervised, similar to how enormous computing infrastructure now runs with less manual oversight, the industry has faced mounting friction between developer productivity and the risk of letting code-executing tools act without a human in the loop.
GitHub has not published the specific rule set its reviewing model uses to judge a request as clearly safe, nor a published rate for how often that judgment is later overridden by a human.
One developer newsletter covering the release noted that the changelog publishes no rule set and no false-approval rate, so treat assisted mode as a probability, not a policy.
What it means for developers
For individual programmers, the practical effect is fewer interruptions during long agent sessions, since routine actions like reading a file or running a standard test command no longer require a manual click every time.
For engineering leaders managing fleets of developers using AI tools, similar in spirit to how companies now monitor sprawling GPU infrastructure at scale, the enterprise controls give administrators a way to permit faster agent workflows without fully removing human review from higher-risk actions.
GitHub’s documentation on the feature, titled “Allowing safe commands to run without manual approval,” outlines further detail on how assisted permissions evaluate requests and when approval is required , according to the company’s changelog entry.
