Copilot Code Review Now Costs Actions Minutes

On April 27, GitHub dropped a changelog entry that every engineering manager running Copilot code review should read: starting June 1, 2026, every Copilot code review on a private repository will consume GitHub Actions minutes. That's in addition to the AI Credits you're already paying for the model calls. Two meters running at once, on every pull request.

If your team reviews 50 PRs a day across a handful of private repos, your Actions minute budget just picked up a new, recurring line item. The good news: you don't have to absorb it at GitHub's rates. Self-hosted runners are explicitly supported, and routing reviews through a provider like Tenki can cut per-review compute costs compared to GitHub-hosted defaults.

What changed and why it matters

Copilot code review isn't a simple API call anymore. GitHub rebuilt it on an agentic tool-calling architecture that pulls in full repository context, runs multi-step analysis, and posts structured review comments. That agent needs somewhere to run. It runs on GitHub Actions runners.

Until June 1, those runner minutes are free. GitHub absorbed the compute cost as part of your Copilot subscription. After June 1, each review draws from your Actions minute entitlement on private repos, with overages billed at standard Actions rates.

Public repositories aren't affected. Actions minutes remain free for public repos. This is purely a private-repo cost.

Two billing dimensions, one review

After June 1, each Copilot code review generates two charges:

1. AI Credits for the Copilot model calls. This is the token-based cost GitHub announced alongside their broader move to usage-based billing. The model does the reasoning; you pay per token consumed.
2. Actions minutes for the runner that hosts the review agent. The agent needs a compute environment to clone, analyze, and interact with your repository. That's an Actions job, billed like any other.

The AI Credits dimension is unavoidable. You can't self-host the model. But the Actions minutes dimension is entirely within your control. GitHub's own docs confirm that Copilot code review supports self-hosted runners and larger GitHub-hosted runners, both billed differently than the standard defaults.

This applies to all Copilot plans: Pro, Pro+, Business, and Enterprise. It also applies to Copilot code reviews from non-licensed users billed via direct org billing.

Estimating your June 1 exposure

How much this costs depends on three variables: how long each review takes, how many reviews your team triggers per day, and which runner type you're using.

Copilot's agentic review typically runs for 3 to 8 minutes per PR, depending on diff size and how many tool-calling loops the agent makes. A team doing 40 reviews a day on the standard Linux 2-core runner ($0.006/min) is looking at roughly:

• Conservative (3 min/review): 40 reviews x 3 min = 120 min/day, or 2,640 min/month. At $0.006/min, that's $15.84/month.
• Moderate (5 min/review): 40 reviews x 5 min = 200 min/day, or 4,400 min/month. That's $26.40/month.
• Heavy (8 min/review): 40 reviews x 8 min = 320 min/day, or 7,040 min/month. That's $42.24/month.

Those numbers look manageable in isolation. But they eat into your existing Actions minute entitlement. A GitHub Team plan includes 3,000 free minutes per month. At 5 minutes per review, 40 reviews a day burns through that entire allowance in under 15 days, just from code reviews. Your CI builds, tests, and deployments still need minutes too.

Enterprise Cloud accounts have 50,000 free minutes, so the impact is proportionally smaller. But most teams on Enterprise are also running far more reviews per day.

The self-hosted runner escape hatch

GitHub's Actions billing docs state it plainly: GitHub Actions usage is free for self-hosted runners. When you route Copilot code review through a self-hosted runner, the review job doesn't count against your Actions minute entitlement at all. You pay the runner provider instead.

GitHub's runner configuration docs for Copilot code review confirm two options: self-hosted runners via ARC (Actions Runner Controller) and larger GitHub-hosted runners. The self-hosted route is the one that removes Actions minutes from the equation entirely.

There are constraints. Copilot code review only supports Ubuntu x64 Linux runners. ARC is the only officially supported self-hosted configuration. And your runners need outbound access to api.githubcopilot.com, uploads.github.com, and user-images.githubusercontent.com alongside the standard GitHub Actions domains.

Routing Copilot reviews through Tenki runners

Tenki Runners are self-hosted GitHub Actions runners priced at $0.002 per core per minute for x64 Linux. Compare that to GitHub's standard 2-core Linux runner at $0.006/min. For a job that requires two cores, Tenki's rate comes out to $0.004/min versus GitHub's $0.006/min.

But the real savings come from what doesn't happen: the review minutes don't draw from your GitHub Actions entitlement. Your 3,000 or 50,000 free minutes stay available for CI builds, test suites, and deployments. No more competing for the same pool.

Configuration is straightforward. In your repository's copilot-setup-steps.yml file, you set the runs-on attribute to your Tenki runner label instead of the default GitHub-hosted runner:

# copilot-setup-steps.yml
jobs:
  copilot-setup-steps:
    runs-on: tenki  # your Tenki runner label
    # ...

That's it. Every Copilot code review that triggers on that repository will now run on Tenki infrastructure at Tenki's per-minute rate. No minute consumption on GitHub's side.

Running the numbers: GitHub-hosted vs. Tenki

Using the moderate scenario from earlier (40 reviews/day, 5 minutes each, 22 working days per month, 2-core Linux runner):

• GitHub-hosted: 4,400 min x $0.006/min = $26.40/month. Plus those 4,400 minutes are gone from your entitlement.
• Tenki Runners: 4,400 min x 2 cores x $0.002/core/min = $17.60/month. Zero minutes consumed from your GitHub entitlement.

The per-minute savings are real but not dramatic. The bigger win is preserving your free minute allotment. For a GitHub Team account, shifting 4,400 review minutes off your entitlement means those 3,000 free minutes go entirely to builds and deploys instead of being split with code review overhead.

Tenki also offers 500 free minutes to get started, so you can test the setup before committing.

What to do before June 1

You have about three weeks. Here's a practical checklist:

1. Audit your current Actions usage. Check your billing settings to see how many minutes you're already consuming. If you're already close to your plan's limit, Copilot reviews will push you into paid overages fast.
2. Count your Copilot reviews. Use GitHub's Copilot usage metrics to see how many code reviews your org runs per week. Multiply by an estimated 3-8 minutes per review to get your projected monthly minute consumption.
3. Set or adjust your Actions budget. GitHub lets you cap spending. Without a budget, overages are unlimited. Set a ceiling that accounts for the new review load.
4. Evaluate self-hosted runners. If the projected cost or entitlement impact is significant, test a self-hosted runner provider for Copilot reviews. Tenki's Starter plan is free with $10 in monthly credits, enough to validate the setup on a real repo.
5. Brief your billing admin. GitHub's own announcement recommends sharing the news with billing administrators. A surprise spike in Actions spend on June 2 is a bad look for everyone.

The bigger picture

This billing change is part of GitHub's broader shift toward usage-based pricing for Copilot. The flat-rate subscription era is ending. AI Credits replace premium request units. Actions minutes get tagged to agentic features that previously ran for free. Each new Copilot capability that uses the agentic architecture will likely carry the same runner cost.

That makes your runner strategy more important than it was a month ago. Every agentic Copilot feature that runs on Actions is a line item you can control by choosing where the job executes. Self-hosted runners aren't just about CI speed anymore. They're about controlling the cost of AI-powered developer tools that GitHub is billing through the Actions meter.

If you want to test Tenki Runners for Copilot code review before June 1, sign up for a free Starter account and swap the runs-on label on one repo. You'll know before the billing change hits whether it works for your team.