If you're building Python AI agents that need to comply with the EU AI Act, this gives you 14 compliance tools right in Claude Desktop, Cursor, or Claude Code. It scans code against six specific articles, detects 15 prompt injection patterns, classifies risk levels, and generates working remediation code with trust layer integrations. The basic install gets you 10 tools for scanning and fixing. Add the full SDK and you unlock GDPR checks, bias analysis, pre-execution validation gates for Article 14, and compliance history tracking. It supports LangChain, CrewAI, AutoGen, and other major frameworks. Under the hood it uses air-trust's HMAC-SHA256 audit chain with Ed25519 signatures. Useful when you need to scan a project directory, validate agent actions before they run, or generate audit reports.
claude mcp add --transport stdio airblackbox-air-blackbox-mcp uvx air-blackbox-mcp