Audison

Multi-agent code auditing. Two LLMs independently verify every finding before it reaches the report — no hallucinated vulnerabilities, no noise.

Run your own audit in one command:

pip install audison[html] audison audit ./your-project --html --output report.html

Results below are pre-generated. Install the CLI and configure your API keys to run live audits.

Example Audit Reports

Click either report to view the full audit with severity breakdowns, arbiter evidence chains, and fix suggestions.

Express.js API Server
A toy Express.js server with deliberately introduced vulnerabilities: raw SQL queries, exposed stack traces, and missing CORS headers.
2 Critical 3 High · 4 findings · 2 arbiters
React Form Component
Client-side form with unsanitized user input rendered via dangerouslySetInnerHTML and missing CSRF token validation.
1 Critical 2 High · 1 finding · 2 arbiters

How It Works

1 Auditor scans your code for 20+ vulnerability patterns
→
2 Arbiter A independently verifies each finding with line-level evidence
→
3 Arbiter B does the same — both must agree
→
4 Report includes only confirmed findings with severity, evidence, and fixes