About
I’m Jose Marques, a developer building open-source tools for AI agent governance.
AI agents are making decisions in production - calling APIs, moving money, accessing data. Most of them run without audit trails, without policy enforcement, and without any way to prove what they did or why.
I built asqav to fix that.
The problem
The EU AI Act requires audit trails and human oversight for high-risk AI systems by August 2026. Most AI agent frameworks have no built-in governance. Adding compliance after the fact is painful.
The solution
asqav is a Python SDK that adds governance to any AI agent in one line of code. It works with LangChain, CrewAI, OpenAI Agents SDK, and MCP servers.
- Tamper-evident audit trails with quantum-safe signatures
- Policy enforcement - block or flag agent actions in real-time
- Token and key revocation
- Human-in-the-loop approval workflows
- CI/CD compliance scanning
Open source
Everything is MIT licensed and on GitHub:
- asqav-sdk - Core Python SDK
- asqav-mcp - MCP server for Claude Desktop
- asqav-compliance - GitHub Action compliance scanner
- eu-ai-act-checklist - Compliance checklist
- asqav-langchain-example - LangChain integration example
- asqav-crewai-example - CrewAI integration example
- asqav-mcp-example - MCP integration example