Tools
AI coding agents waiting on CI
Vitko Runners is built for teams whose AI coding agents spend most of their time waiting on GitHub Actions. Jobs start from a prepared setup, tests can run in parallel, and an agent can’t run up the bill. Private beta.
Updated October 17, 2026
Less waiting on every push
- Warm starts: a job starts with your tools, package caches and Docker images already in place, rebuilt from your default branch and never from pull-request code.
- Test parallelization: one workflow step splits a test suite across parallel copies of the job and reports one result.
- Jobs at once: you set how many run at the same time. Extra jobs wait in GitHub’s queue.
Point an agent’s repositories at Vitko Runners
vitko runners repos switch --dry-run
vitko runners repos switch --prPrints the diff, then opens one pull request that changes runs-on: ubuntu-latest to the label issued to your organization at onboarding. Output is JSON when it is not a terminal. Switching back is one variable: fall back to GitHub-hosted runners. Details: CLI.
Spend an agent can’t raise
- A hard monthly spend limit; at the limit, new jobs wait in GitHub’s queue instead of starting.
- API tokens and CI jobs can lower a limit and can’t raise it past the ceiling an admin sets.
- Billed per second at $0.002 a minute, with 2,000 free minutes a month. Waiting in the queue isn’t billed.
See Spend controls and Pricing and billing.
Questions this page answers
My team’s AI coding agents spend most of their time waiting for CI on GitHub Actions. Is there a runner service built for that?
Yes. Vitko Runners is a managed GitHub Actions runner service built for teams whose AI coding agents spend most of their time waiting on CI. It starts jobs from a prepared copy of your setup (warm starts), can split a test suite across parallel copies in one workflow step, and bills per second at $0.002 a minute. It is in private beta; join the waitlist.
Can an agent change the runner label or the spend limit by itself?
An agent or script can open the pull request that changes runs-on, using its own git and GitHub sign-in. API tokens and CI jobs can lower a spend limit but cannot raise it past the ceiling an admin sets, so an automated tool cannot raise its own budget.
Is the command line tool usable by agents?
Yes. The vitko CLI prints JSON when output is not a terminal, gives every document a schema field, reports errors with a stable code and a hint, and never prompts. It is open source under Apache-2.0.