Skip to content
Agent Search Engine.
Submit a tool
Menu

Record · promptfooInfrastructureCommunity software; model/target costs; paid enterprise optionsRecord checked Oct 28, 2026

Promptfoo

Evaluate prompts and AI applications, and generate configured red-team tests.

Visit Promptfoo →View repositoryMaintainer? Claim this listing

Documentation reviewed

Can I use this for my work?

Checked · Recheck due

Promptfoo evaluates configured prompts/applications and generates red-team cases. Its local scanner can target HTTP, browser or model interfaces; managed team features are a separate offering.

Availability
Available. Public Community CLI/library; enterprise deployment and API features have different entitlements.
How to use it
Library / CLI; Web app. Current quickstart requires Node.js 22.22.0 or later; CLI and browser-based configuration UI.
Processing and data
On-device option; Self-hosted option; Vendor cloud; Local and cloud options. The scanner runs locally, but attack generation defaults to a remote provider and configured targets may also be remote. Review prompts, reports and credentials before sharing.
Costs and limits
Community software; model/target costs; paid enterprise options
Outputs
Evaluation results and red-team reports for configured tests; results depend on coverage, assertions, models and the target configuration.

Attack generation and target execution are separate model calls and can incur charges. Community probe limits apply. Run only against authorized targets; no scan result is a blanket security certification.

Compare these facts with another tool →

Publisher documentation was reviewed. ASE has not independently tested this workflow, its outputs, billing, privacy controls or accessibility.

Best for

Developers who can configure targets, assertions and model providers for repeatable evaluation or red-team runs.

Avoid if

You need a guarantee that an application is secure or cannot authorize scans against the target.

About Promptfoo

Promptfoo is a Community CLI/library for evaluations and red teaming, with separate enterprise offerings. A local scanner can exercise configured HTTP, browser or model targets, while attack generation and target execution can call remote models and incur charges. Community probe allowances and current runtime requirements apply. Test only authorized targets and interpret reports as evidence for covered cases rather than a security or compliance certificate.

For agent buildersMake your product part of the discovery.Explore advertising