Technology
Open-source Conduct tool released for AI governance and tool call security
The project includes a policy engine, LLM proxy, and cryptographic audit log available under the Apache 2.0 license.
The short version
- Conduct has launched as an open-source governance system to evaluate and filter LLM requests, CLI hooks, and Model Context Protocol (MCP) tool calls.
- The software introduces pre-execution policy checks, workspace-signed configurations, and a SHA-256 hash-chained audit log to prevent unauthorized AI actions.
- Teams can deploy the tool locally using Docker Compose or Kubernetes, with a hosted commercial version also available.
Key facts
- The open-source repository includes Conduct Guard for policy decisions and Conduct Router as an LLM proxy under the Apache License 2.0.[Hacker News]
- Conduct enforces rules across LLM requests, CLI tool hooks, and MCP invocations prior to action execution.[Hacker News]
- The release ships with over 20 compliance packs addressing frameworks such as OWASP, SOC 2, HIPAA, PCI DSS, NIST AI RMF, and the EU AI Act.[Hacker News]
- Audit trails are maintained using a SHA-256 hash chain rooted at workspace creation to allow independent verification of log entries.[Hacker News]