GITHUB ACTIONS COSTS
Your CI spend. Your potential savings.
Compare your current CI costs with alternative scenarios.
Savings are not guaranteed. Potential savings are estimates based on assumptions. CI Optimizer does not guarantee lower costs. Actual costs may differ.
Start with your own data
Choose what you want to understand. You can use each file type on its own.
What did my CI cost?
Use a GitHub metered usage report to see Actions spending before and after discounts, and compare periods.
- Open billing for the personal account or organization that pays for the usage. Go to Usage / Metered usage.
- Choose Get usage report, select your dates and request the report by email. Save the CSV from the download link.
- Enter that account or organization here, then select the CSV files together.
USD only. Up to 12 files for the same account with non-overlapping charge-date ranges. A billing CSV alone cannot estimate savings from different CI rules.
What could I change?
Use a prepared CI Optimizer history file for one repository to review repeated work and model alternative CI rules or runners.
Ask the person preparing your CI data for a CI Optimizer snapshot or a compatible GitHub runs-and-jobs export. You can also collect existing history from a connected repository in your CI workspace.
The file needs run attempts and their jobs, not just a list of runs. Export JSON reports from this app are analysis results and cannot be imported as history. Up to 5 MiB, 1,000 attempts and 20,000 jobs.
View sample input format (opens in a new tab) · fictional data
How is my workflow set up?
Save a .yml or .yaml file from your repository's .github/workflows/ folder.
Open Workflow review and import that file to inspect triggers, runners, concurrency and cache declarations with line references.
One UTF-8 file up to 128 KiB. YAML alone cannot establish spending, runner compatibility or safe changes. The review does not execute or edit your workflow.
Open workflow review (opens in a new tab)Using history and billing together? Import history first, then billing for the same account. Confirm prices and scenario assumptions before calculating. A lower bill after a change does not prove the change caused the reduction.
Imported data stays in the current tab's memory. Refreshing or clearing that tab removes it. Workflow review opens in a separate tab, so your spending analysis stays here. Reports can be viewed in the app; choose JSON in the report preview to copy the full results.
Potential savings over time
Illustrative example · USDSpending & savings
Waiting for dataImport spending from GitHub billing
Select one or more GitHub usage CSV reports for the same account to see spending after discounts. Each selection replaces the previous billing import. Files stay in this browser.
USD reports only: up to 12 files, 5 MiB each, 10 MiB and 20,000 rows in total. Included charge-date ranges must not overlap, even across different repositories. Only Actions minute charges are included; storage, tax and other products are excluded. Billing alone cannot show what different CI rules would have saved.
Compare spending before and after a change
Choose the period covered by your reports and the first day after your change. Compare reported net spending by day, week and month.
What could your CI have cost?
Import run history, choose an optimization scenario and confirm your assumptions to estimate savings over the imported period.
Try the September example to compare reported spending by day, week and month, and see historical macOS and Linux prices.
Download synthetic run history for the rule simulatorEstimate your savings
Based on imported runsCompare the same imported work with one selected scenario. This does not change your workflows.
Save one calculated estimate, then change the proposed runner or rule and calculate again to compare two options.
Two scenario estimates
Saved in this tab only. Refreshing, importing new data or changing original prices or the billing model clears the saved scenario.
Run details & supporting minutes
Time after a newer PR trigger. Review required.
Time by workflow
Sum of completed job intervalsReview candidates
Suggestions, not automatic changesCoverage and calculation limits
Parallel jobs and repeat attempts are counted separately. Findings do not prove equivalent tests or achievable savings.