Documentation

AgentFuzz.AI is an AI vulnerability assessment and assurance platform that helps you identify vulnerabilities in your AI systems using industry frameworks — OWASP LLM Top 10 and NIST AI RMF. This documentation covers everything you need to get started and use the platform effectively.

Getting Started

Quick start guide — sign up, create a project, run your first scenario in 5 minutes.

Projects

Configure AI targets for assessment. API endpoints or chat widgets with XPath selectors.

Prompt Library

Browse 500+ attack prompts mapped to OWASP and NIST. Add custom prompts for your org.

Scenarios

Create assessment scenarios, execute them, track progress, and manage run history.

Reports & Scorecards

Understand pass/fail reports, security scorecards, and compliance coverage gaps.

Organizations

Manage teams, members, roles, and org-scoped resources like custom prompts.

Key Concepts

Prompt — An assessment input designed to probe a specific vulnerability. Prompts are mapped to OWASP or NIST categories and include expected behavior and risk descriptions.

Organization — A team or company. Projects, members, and custom prompts are scoped to an organization.

Project — An AI system you want to assess. Each project has a target URL and a type (API endpoint or chat widget).

Scenario — A collection of prompts grouped into a runnable assessment suite. You select prompts from the library and execute them against a project.

Run — A single execution of a scenario. Each run produces results (pass/fail) for every prompt. You can run a scenario multiple times and compare results.

Scorecard — An aggregated compliance view showing pass rates across OWASP and NIST categories. Identifies coverage gaps where categories haven't been assessed.