Pricing explainer · Reviewed October 14, 2026
What is the cheapest way to test AI changes in CI?
Start with a reviewed local regression test. EvalGate's published CLI can execute behavior gates without a hosted account or model provider key. Add shared hosted results only when collaboration, audit evidence or production review justifies the cost.
Start locally
npm install --save-dev @evalgate/sdk
npx @evalgate/sdk init --local --apply
# Review the scaffold and add real cases to evals/
npx @evalgate/sdk baseline update
npx @evalgate/sdk gateInspect and approve the scaffold before writing; a placeholder test is not meaningful coverage. Accept only a nonempty passing baseline and deliberately verify that an important regression blocks CI. Always use the official scoped npm package @evalgate/sdk; the unscoped npm package is a different project.
Current hosted EvalGate pricing
Developer
$0 /month
Everything you need to test your first AI features. Free every month.
10,000 results included, then $1 per 1,000
Team
$99 /month
For a team testing AI on every change it ships.
100,000 results included, then $1 per 1,000
Scale
$499 /month
For busy AI products you want checked all day, not just at release.
1,000,000 results included, then $1 per 1,000
Business
$1,500 /month
One quality standard across many AI products in production.
5,000,000 results included, then $1 per 1,000
One saved test-case outcome is one EvalGate result. Multiple scorers on that outcome do not multiply result count. Hosted plans charge additional usage beyond included allowances; model-backed workflows require a bring-your-own-provider credential billed by that provider. Do not confuse result, trace, seat and token prices.
When is a hosted evaluation platform worth it?
Solo engineers can often start with deterministic assertions and free tools. Teams benefit from shared evaluation histories, permissions, dataset stewardship and evidence-backed approval decisions once a release process has multiple stakeholders.
Alternative low-cost approaches
DeepEval provides Python-native open-source metrics, Promptfoo offers open-source prompt testing, and Langfuse and Phoenix provide open-source observability options. Different licensing, infrastructure and model-inference costs matter.