About Vigile

The trust layer for AI agents.

The Problem

"AI agents are connecting to tools they've never verified."

AI agents are becoming the new software supply chain. MCP servers, agent skills, and tool definitions are installed from community registries with little to no security review. A malicious MCP server can exfiltrate sensitive data, inject hidden instructions, or establish covert communication channels — all while appearing perfectly legitimate.

220+detection checks across Vigile
0-dayconfig needed to get started
100%open-source CLI, Apache 2.0

Our Approach

Static Scanning

Multi-engine scanner with 220+ detection checks across MCP, skill, runtime, and backend analysis.

Trust Scoring

5-factor weighted model (code analysis, dependency health, permission safety, behavioral stability, transparency) producing a 0-100 trust score.

Runtime Monitoring

Sentinel watches what MCP servers actually do on the network — detecting C2 beaconing, data exfiltration, DNS tunneling, and credential theft in real time.

Community Registry

Crowdsourced trust database for MCP servers and agent skills, with automated crawling from npm and GitHub.

The Solution

Vigile wraps every layer of the AI agent supply chain — from static source analysis and dependency scoring to real-time network monitoring — into a single trust signal you can act on. One command, zero config.

terminal
$npx vigile-scan @anthropic/mcp-server-github
▸ Scanning @anthropic/mcp-server-github@1.2.0...
✓ Fetching source from npm registry
✓ Running 220+ detection checks
✓ Evaluating 5-factor trust model
Trust Score 92 / 100
Risk Level LOW
Findings: 0 critical 0 high 1 info
INFO README.md lacks permission disclosure
✓ Scan complete — no threats detected

By the Numbers

12,400+Servers indexed
340K+Scans run
1,800+Threats detected
620+CVEs referenced

Open Source & Transparency

The Vigile CLI scanner is free and open source (Apache 2.0), available on npm. Every trust score is computed against a documented, public methodology — no black boxes. Our security policy is public, and we welcome responsible disclosure from the community.