Skip to main content

VORO

The trust layer for the autonomous agent economy

Local-first security scanning for smart contracts and AI agents.
745 patterns. 16 languages. 14 external scanners. Nothing leaves your machine.

~

Local-First

Your code never leaves your machine. No cloud uploads, no API keys required, no vendor lock-in. The full detection pipeline runs locally.

|||

Three Detection Layers

Static analysis (voro-scan) feeds into Bayesian scoring (voro-brain) for confidence-weighted A-F grades. Findings confirmed by multiple methods score higher.

>_

Agentic Security

MCP server trust boundaries, prompt injection vectors, unconstrained tool permissions. VORO covers autonomous agent threats that no other scanner addresses systematically.

%

Published Accuracy

Slither precision: 0.746. SmartBugs recall: 0.972. Every number comes from a labeled corpus with reproducible evaluation. No competitor publishes these metrics.

{ }

16 Languages

Solidity, Python, JavaScript, TypeScript, Go, Rust, Move, Vyper, and more. Auto-detection — point it at a repo and it runs the right checks.

>>

Open Source

The scanner is MIT-licensed. 745 active patterns, 14 external scanner integrations, 9 taxonomy mappings. Free for individuals and commercial use.

Detection Pipeline

1

Scan

agent-builder audit . runs 745 patterns + 14 external scanners against your codebase. Output: structured JSON findings.

2

Score

voro-brain evaluates findings across 6 risk dimensions using corpus-calibrated Bayesian priors. Output: A-F safety grade.

3

Report

View results in the web UI, as GitHub PR comments, or consume the JSON directly in your CI pipeline.

Start scanning in 5 minutes

pip install agent-builder && agent-builder audit .
Read the Quick Start