Promptfoo Probe-Based Pricing Model

Probe based pricing turns Promptfoo from a seat sold to a security team into a metered testing utility that grows with every scan. That matters because the expensive part of the product is not the dashboard, it is generating attack prompts and grading results with model inference. A company testing one chatbot before launch and a bank scanning dozens of live agents every week create very different compute loads, and probe pricing tracks that difference closely.

• A probe is one request against the target system during red team testing, and the free tier includes 10,000 probes per month. Enterprise plans add governance features like SSO, RBAC, reporting, API access, and on prem deployment, then expand spend through extra probe capacity as scan volume rises.
  1 promptfoo 2 sacra
• This model fits the actual workflow. Promptfoo runs automated attacks across many vulnerability types, then grades outputs and generates remediation guidance. More apps, more attack scenarios, and more frequent scans all mean more requests, so usage is a cleaner unit than charging per security seat.
  2 sacra
• The same structure also protects margins better than flat pricing. Competitors like Protect AI also run large attack libraries, while cloud tools like AWS Bedrock Guardrails are built into existing platforms and can cover lighter use cases. Metering helps preserve pricing power when inference heavy testing would otherwise be underpriced.
  2 sacra 3 protectai 4 amazon 5 amazon

The next step is continuous security, not one time testing. As AI teams move from pre release checks to always running scans and runtime controls across fleets of agents, probe volume should become a larger share of contract value, and the winning vendors will be the ones that tie each extra dollar of spend to a clearly visible increase in coverage.