GitHub Copilot Automations: Schedule Your AI Agent

GitHub launched Automations in public preview, a feature that lets developers schedule the GitHub Copilot cloud agent to run recurring tasks on a repository without manual prompts. Each Automation attaches a natural-language prompt, a schedule or event trigger (issue created, PR opened, comment posted), and an explicit tool permission list to a repository, running in an ephemeral GitHub Actions VM on GitHub's infrastructure. GitHub's official documentation ships three reference use cases — automatic issue labeling, a nightly test fix that opens a draft pull request, and weekly release notes — and as of August 2026 Automations also support comment triggers.

GitHub Copilot just stopped being something you talk to and started being something that runs. The Automations feature — now in public preview — lets you schedule the cloud agent to work on a recurring basis, no manual prompt required. Set it up once, define the trigger, and Copilot handles the rest: fixing failing tests overnight, labeling issues as they land, drafting release notes before your Friday standup. This is not a chatbot improvement. It is infrastructure. Automations vs. Regular Agent Use: What Actually Changed Until now, Copilot’s coding agent was interactive. You open an issue, assign it to Copilot, and wait for a pull request. Useful — but still a manual step. Automations remove that step entirely. An Automation is a persistent configuration attached to a repository. You define a natural-language prompt describing the task, select a trigger, and grant the agent specific tools it is allowed to use. From that point on, it runs automatically — in an ephemeral GitHub Actions VM, on GitHub’s infrastructure, completely independent of your machine. The mental model is the same as how CI/CD changed testing. You did not stop running tests; you stopped triggering them manually. Automations do the same thing for AI-driven tasks. The pipeline runs. You review the output. How to Set Up Your First Automation There are two paths to create an Automation. On github.com , navigate to the Agents tab in your repository, select Automations, and click Create new. In the GitHub Copilot desktop app , find Automations in the left sidebar, click New automation, and enable the “Run as cloud automation” toggle. Either way, you configure four things: - Name: A human-readable label e.g., “Nightly test fixer” - Prompt: Natural language instruction for the agent - Trigger: When it runs — schedule hourly, daily, weekly or event issue created, PR opened, comment posted - Tools: Explicit permissions — what the agent is actually allowed to do create pull request, update issue labels, etc. The tools list is not a formality. It is a hard permission boundary. An agent without “create pull request” in its tool list cannot open a pull request, regardless of what the prompt says. Design this deliberately. Three Use Cases Worth Setting Up Today GitHub ships three reference use cases in its official Automations documentation https://docs.github.com/en/copilot/concepts/agents/cloud-agent/about-automations . These are a reasonable starting floor. 1. Automatic Issue Labeling Trigger: Issue created Prompt: "Automatically label new issues as bug, enhancement, or question based on their content." Tools: update issue labels Every time a new issue lands in your repository, the agent reads it and applies an appropriate label. No triage ceremony required. Particularly useful for open-source projects and any team that struggles to keep the issue tracker organized. 2. Nightly Test Fix Trigger: Daily schedule Prompt: "Check for failing tests on the main branch, attempt a fix, and open a draft pull request." Tools: create pull request, read repository This one runs while you sleep. If the agent finds a failing test and can fix it, you wake up to a draft PR. If it cannot, nothing happens — no noise, no false alarm. It is an on-call engineer who never sleeps and never bills overtime, though cloud sandbox compute time is metered more on that below . 3. Weekly Release Notes Trigger: Weekly schedule Prompt: "Review merged pull requests from the past week, draft release notes summarizing changes, and open a pull request." Tools: create pull request, read repository Release notes are one of those tasks everyone agrees should happen and nobody wants to do. This Automation does not need to be perfect — it needs to give the engineer a starting point. That cuts the task from thirty minutes to five. As of August 2026, Automations also support comment triggers https://github.blog/changelog/2026-08-03-trigger-copilot-automations-with-comments/ . Add @copilot in an issue or PR comment with instructions, and it fires. This turns any team member into a trigger — no admin access required. Availability and Cost: The Numbers You Need Automations are available on Copilot Pro $10/month , Pro+ $39/month , Max $100/month , Business $19/user/month , and Enterprise $39/user/month . The Free tier is excluded. Business and Enterprise accounts require an administrator to enable the cloud agent policy first. The billing detail that matters: cloud sandbox usage is charged at $0.000024 per compute second . A nightly test-fix run spending ten minutes in the sandbox costs roughly $0.014. Run it every night for a month — about forty cents. Multiply by a dozen automations across multiple repositories and the math still works. But model token usage stacks on top of that. Monitor your usage dashboard once you scale up. Why This Is Different From What Came Before No other major AI coding tool offers this. Cursor https://cursor.com runs in your IDE. Claude Code is interactive. OpenAI’s Codex handles task assignment but has no built-in scheduling. GitHub Copilot Automations is the first to treat the AI coding agent as scheduled infrastructure — running inside the repository layer, governed by the same policies that govern your Actions workflows. That integration matters. Automations live where your code lives, operate with your repository’s existing identity and governance controls, and produce output in the format your team already reviews: pull requests and issue updates. There is no new workflow to adopt. There is just less manual work. The public preview label means things will change. Prompts will need tuning. Some runs will fail or produce mediocre output. That is expected. Start with issue labeling or the nightly test fix — low risk, immediate value, and a good way to understand how your agent behaves unsupervised before you give it anything more sensitive. Full setup instructions are in GitHub’s documentation https://docs.github.com/en/copilot/how-tos/use-copilot-agents/cloud-agent/create-automations . The public preview announcement is on the GitHub Changelog https://github.blog/changelog/2026-09-25-github-copilot-weekly-releases-september-21/ .