100% Client-Side SAST · Air-Gapped Browser Execution · OWASP LLM06 & LLM01
MCP & Autonomous AI Agent Security Scanner
Audit your Model Context Protocol (MCP) server configurations (`mcp.json`), tool schemas, Docker/agent manifests, and system prompts against privilege escalation, unrestricted shell execution, credential leakage, and indirect prompt injection vectors. Zero telemetry: analysis runs entirely in your local browser memory.
1. Paste Config (mcp.json / Tool Schema / Prompt)
2. Static Security Audit Findings
10 Active Rules
CVSS-AI Exposure Rating
Ready to Scan
0.0 / 10
Click "Load Vulnerable Sample" then "Run Adversarial Audit" to inspect a typical insecure MCP agent configuration.