Tool overview
Fabraix is listed under AI Governance Safety & Evaluation AI tools.
What is Fabraix?
Fabraix builds Nyx for autonomous black-box adversarial testing and Arx for runtime defense of AI agents. It tests security, logic, alignment, prompt injection, tool use, and multi-turn failure modes.
Best for
Adversarial testing and runtime defense for AI agents
Who is it for?
Decision note
Rechecked the product, pricing, documentation, security, terms, and current Y Combinator profile. Nyx black-box red teaming, Arx runtime defense, APIs, CI/CD use, a free application-based Research tier, custom Team/Enterprise plans, Google Cloud hosting, configurable retention, Fabraix, Inc., founders@fabraix.com, and 2026 founding were confirmed. A public free trial and affiliate program were not confirmed.
Key features
Runs autonomous black-box red-team scans
Tests security, logic, and alignment weaknesses
Uses more than 1,000 adversarial strategies
Supports CI/CD retesting and continuous coverage
Provides findings dashboards and remediation support
Offers runtime defense APIs through Arx
Use cases
Red-team customer-facing AI agents
Test browser, voice, and multi-agent systems
Measure prompt-injection and tool-use risk
Retest agents after prompt or model changes
Add runtime checks for dangerous actions
Pros
- Requires no source code or model weights for Nyx
- Publishes transparent pricing tiers
- Offers both testing and runtime defense
- Grounded in original adversarial-security research
Limitations
Continuous testing can add operational cost. Confirm current product scope, security, retention, and commercial terms before deployment.
Pricing details
Billing options
Pricing note
Researchers can apply for $0 non-commercial Nyx access. One-off scans are custom-priced per agent, and continuous coverage is custom-priced monthly by usage. No separate time-limited free trial is published.
Supported languages
- English
Integrations
Agent endpoints
URLs
Phone agents
CI/CD
Nyx API
Arx API




Please log in to join the discussion.