Copilot Automations Guide: Shared, Cloud, Self-Improving

Last updated: October 31, 2026

Copilot Automations run GitHub Copilot cloud agent on a schedule, on repository events or through the API. GitHub’s October 28, 2026 updates added shared team automations that act as Copilot, cloud automations in VS Code 1.144 that run while your computer is off, and self-improving automation groups that propose prompt changes for your review.

This guide draws on GitHub’s Copilot Automations documentation, three October 28 changelog posts and the VS Code automations documentation. If you are still choosing a plan, see our comparison of Copilot Pro vs Pro+.

Contents

What is new in October 2026
Personal vs shared Copilot Automations
How to create a Copilot automation
Triggers, tools and limits
Self-improving automation groups
Practical workflows
Limitations
Frequently asked questions

What is new in October 2026

According to the documentation, an automation is a saved Copilot cloud agent task with a name, a type, a prompt, one or more triggers and a set of allowed tools. You can optionally pick the model and reasoning level and set AI credit limits. GitHub shipped three changes on October 28, 2026:

Shared automations. Per the shared automations changelog, automations were previously personal only. Shared automations now belong to a repository, act as Copilot and are billed to the repository owner. They are rolling out to Pro, Pro+, Max, Business and Enterprise users.
Cloud automations in VS Code. The VS Code changelog post says automations created in the Agents window can now run in GitHub’s cloud, including on GitHub events, instead of only locally on a schedule.
Self-improvement. The self-improvement changelog describes automations that learn from outcomes, such as whether people change labels or merge pull requests, and propose prompt refinements.

Personal vs shared Copilot Automations

You pick the type when you create an automation, and you cannot change it later. The table below summarizes GitHub’s comparison.

Setting	Personal	Shared
Who can see it	Only you	Everyone with read access to the repository
Who can see its runs	Everyone with read access	Everyone with read access
Who can edit it	Only you	Everyone with write access
Who pays for AI credits	You	The organization or user that owns the repository
Identity on GitHub	Your identity	Copilot’s identity
If the creator leaves	Tied to the creator	Keeps running, with change history

For a shared automation in an organization-owned repository, an administrator must enable the “Allow use of Copilot cloud agent billed to the organization” policy.

How to create a Copilot automation

You can create automations from the Agents tab of a repository on GitHub, the Automations tab in the GitHub Copilot app, the Agents window in VS Code, or the REST API. In VS Code, the cloud steps are:

Check eligibility. Use a private or internal repository where you have write access and Copilot cloud agent is enabled. Public repositories are not supported.
Update VS Code. Install version 1.144 or later and confirm chat.automations.enabled and chat.automations.cloud.enabled are on.
Open Automations. In the Agents window sidebar, select Automations, then create a new automation or start from a template.
Turn on Run in the Cloud and choose the repository. The cloud run does not use your local working tree or uncommitted changes.
Write the prompt with the task, scope, expected output and safety rules.
Grant tools. Select Configure allowed tools. No tools are selected by default, so add only what the task needs.
Choose an owner. Only me creates a personal automation; Repository creates a shared one that runs as Copilot.
Set the trigger (manual, a recurring schedule or repository events), review the settings and select Create. Test it once with Run now.

Triggers, tools and limits

Triggers. GitHub lists hourly, daily or weekly schedules, issue created, pull request opened and pull request synchronized. Event triggers accept a search query filter, and pull request triggers can also filter by changed files. Every automation can be started through the API with extra context passed into the prompt.

Tools. GitHub tools act only in the one repository the automation is scoped to. Tools from MCP servers configured in the repository can reach whatever those servers’ credentials allow, so review them before granting access.

Inherited settings. Runs use the repository’s Copilot cloud agent configuration: custom instructions, agent skills, firewall rules, and secrets and variables. Automations are stored outside the repository and are not committed to Git.

AI credit limit	Behavior
Per run	Stops the run when reached; minimum 30 AI credits
Daily	Blocks new runs; resets at 00:00 UTC
Weekly	Blocks new runs; resets Mondays at 00:00 UTC
Monthly	Blocks new runs; resets on the 1st at 00:00 UTC

Each run uses GitHub AI Credits and GitHub Actions minutes, and Actions minutes are billed to the repository owner. Daily, weekly and monthly limits do not stop a run already in progress, so set a per-run limit to cap each run. For plan allowances, see our answer on Copilot Free limits.

Self-improving automation groups

Automations can be grouped so they share context. When you ask Copilot to set up a self-improving automation, it can create a worker automation plus an evaluator. GitHub’s example is a review automation that comments on new pull requests and a weekly improvement automation that studies how developers responded and proposes refinements.

Proposals can change prompts, tools or triggers. GitHub notifies you, shows a description of the change, and lets you apply or dismiss it.

Practical workflows

Bug reproduction: trigger on issue created, filtered to a bug label, and allow only comment tools so Copilot reports steps without pushing code.
Review screenshots: trigger on pull request opened with a changed-files filter for UI paths, and allow attachment uploads.
Custom reviewer: pair a review automation with a weekly evaluator so prompts improve from accepted and rejected comments.

If you also route work across models, Copilot Automations pair with GitHub Copilot Fusion. For comparable agent tooling from other vendors, see our guides to the Cursor Agent SDK, Claude Code and the Microsoft Copilot harness.

Limitations

Automations run only in private and internal repositories, not public ones or repositories owned by managed user accounts.
Shared automations are still rolling out, and VS Code cloud automations are in preview.
By default, automations ignore events from users without write access, to reduce prompt injection risk. You can opt in to those events, but GitHub flags the risk.
Pull requests from a shared automation need an extra approval by default when the repository requires one, and Actions workflows on automation pull requests wait for a user with write access to approve them.
Session logs, prompts and changes are visible to everyone with read access. To keep an automation as code and review it through pull requests, GitHub points to GitHub Agentic Workflows.

Frequently asked questions

What are Copilot Automations?

Copilot Automations let GitHub Copilot cloud agent run automatically on a schedule, when a repository event happens, or when called through the REST API. You write a prompt once, choose triggers and allowed tools, and Copilot runs the task in the cloud whenever a trigger fires.

Which plans include Copilot Automations?

GitHub lists Copilot Automations for the Copilot Pro, Pro+, Max, Business and Enterprise plans. Automations work only in private and internal repositories, and Copilot cloud agent must be enabled for the repository.

What is the difference between personal and shared automations?

A personal automation is visible and editable only by its creator, acts as that user and bills that user’s AI credits. A shared automation belongs to the repository, can be edited by anyone with write access, acts as Copilot and bills the repository owner.

How do I turn on cloud automations in VS Code?

Update to Visual Studio Code 1.144 or later and enable the chat.automations.cloud.enabled setting. Then open the Automations tab in the Agents window, create an automation and turn on Run in the Cloud.

How much do Copilot Automations cost?

Each run starts a Copilot cloud agent session that uses GitHub AI Credits and GitHub Actions minutes. You can cap usage with optional per-run, daily, weekly and monthly AI credit limits; the per-run minimum is 30 AI credits.

Can Copilot Automations change themselves?

They can propose changes, but not apply them on their own. Self-improving automations use outcome signals on GitHub to suggest edits to prompts, tools or triggers, and you review and accept or dismiss each proposal.

Sources

Image: GitHub mark (Octicons), by GitHub, MIT license, via Wikimedia Commons. Placed on a white background and resized. GitHub and Copilot are trademarks of GitHub, Inc. FreeAnswers is not affiliated with or endorsed by GitHub.

Copilot Automations: Power-User Guide to Shared, Cloud and Self-Improving Workflows