Bernstein – open-source governance layer for AI agents (bernstein.run)

Bernstein has launched an innovative open-source governance layer for AI agents, designed to enforce policies and ensure deterministic execution of code without any model intervention. By utilizing a Python scheduler, Bernstein runs AI agents in parallel while generating a verifiable audit trail for each execution. The latest version, 3.14.159, introduces enhancements such as improved error reporting, a new Muse Code adapter, and a system for signing run receipts, boosting usability and accountability for developers. This project is significant for the AI/ML community as it brings a robust framework for orchestrating AI coding agents like Claude Code and Codex, promoting transparency and trust in automated processes. With features like HMAC-signed audit logs that detail every step of an agent's run, Bernstein sets a new standard for traceability. The integration with Cloudflare enhances its functionality by enabling the management of AI workflows in cloud environments, furthering its applicability for real-world applications. This governance layer offers developers not just tools for AI execution but also an assurance of ethical and verifiable AI operations.