Every time you use AI,it costs money.
Know what a change costs,before you merge it.
- whatcost / cost-regression (pull_request)RequiredFailing after 3s
- build (pull_request)Successful in 1m 12s
- test (pull_request)Successful in 48s
- typecheck (pull_request)Successful in 22s
+$226.83a month
One model swap, in one pull request. This feature goes from $66.62 to $293.45 a month, 340.5% up - and the check ran before anyone merged it.
Use it where you already work
The pricing engine over MCP, so Claude Code, Cursor or whatever agent you already use can price a change before you write it. No key, no model calls of its own.
Settings, then Connectors, then add a custom connector:
In a terminal agent instead?
Diff a change
Point it at the branch. See what the change does to the bill before you merge.
+$226.83 / mo (340.5%) · over budget · exit 1
Plan the cost
Describe the AI feature in a plain YAML file. Get the monthly bill, itemized by step, before you build it.
TOTAL $8.67 / mo · 4 steps priced
Swap the model
Try another model and watch the bill move. Nothing else about your pipeline changes.
TOTAL $55.88 / mo (was $8.67)
Re-cost it at scale
Change one number and see what the same feature costs at production traffic.
TOTAL $86.59 / mo
Sweep the scale
See where the cost curve bends before you commit to an architecture.
10k $8.67 · 100k $86.59 · 1M $865.76
Wire it into CI
Machine-readable output for CI checks, budget gates and spreadsheets.
{ "monthly": 8.67, "low": 7.95, "high": 9.39 }
Pipelines are plain YAML. Point your editor at whatcost.dev/schema/pipeline.json for validation and autocomplete.
Prefer a UI? Open a pipeline in the browser, same engine, nothing to install.