AI Costs
How AI billing works in QAID, ballpark monthly estimates, and tips for keeping costs predictable.
QAID's AI features (test generation, failure classification, page understanding) run through Anthropic's Claude API. You bring your own API key — usage shows up on your own Anthropic invoice with no markup from QAID.
How billing works
You pay Anthropic directly for AI calls. QAID never sits in the middle of your AI bill. This means:
- No markup — you pay Anthropic's published per-token rate
- No metering surprises — you set your own usage caps and budgets in the Anthropic console
- No vendor lock-in — you can rotate keys or change providers without touching QAID licensing
Typical monthly spend
Most teams settle into one of these ranges, depending on test volume and run frequency:
| Team profile | Typical AI spend |
|---|---|
| Light (small project, low run frequency) | ~$30–80/mo |
| Standard (one product, regular CI runs) | ~$100–400/mo |
| High-velocity (multiple projects or many deploys/day) | ~$500–1,500/mo |
| Heavy (large suite + many parallel runs) | $1,500+/mo |
These are estimates only — your actual spend depends on test count, run frequency, page complexity, and how often you regenerate scenarios.
Where the spend goes
The two biggest cost drivers, in order:
- Test execution — running scenario tests is the largest line item for most teams. Cost scales with
tests × runs per day. - Test generation — generating new test scenarios from a crawl is a one-time per-project cost when you onboard, then occasional as you add features.
Other features (page understanding, failure classification, bug detection) are small contributors.
Keep your bill predictable
A few practical tips:
- Set a per-API-key budget in the Anthropic console (Settings → API Keys → Budgets). Anthropic will alert you and pause the key if you exceed it.
- Use the API Usage dashboard in QAID (Admin → API Usage) to see daily/weekly trends per project.
- Right-size your test cadence — for marketing or low-criticality pages, hourly CI runs probably aren't needed.
- Regenerate scenarios sparingly — generation is the one-time bulk spend per project. Once tests are stable, you don't need to re-generate.
See your actual spend
In QAID: Admin → API Usage shows historical cost breakdowns per project. On Anthropic's side: console.anthropic.com → Usage → invoice history.
If your bill ever looks off, check both dashboards — they should match within rounding.