Asqav AI Agent Governance

A governance layer for AI agents that runs pre-execution policy checks and produces signed audit trails. Exposes MCP tools for gating actions before they run, enforcing tool policies with rate limits and approval rules, and creating FIPS 204 ML-DSA signed receipts that bind prompts, tool calls, and outputs together. Includes a tool definition scanner that flags prompt injection, hidden Unicode, and typosquatted tool names before you trust an MCP server. Integrates with Claude Desktop, Claude Code, and Cursor over stdio. The free tier hashes action context locally and stores only metadata server-side. Point ASQAV_API_URL at your own deployment if you want full context for PII redaction or richer policy logic. Useful when you need forensic proof of what an agent attempted or a kill switch that works before the damage happens.