Cryptographic provenance for AI agents from action capture to audit reports (github.com)

RootSign, a product of Providex AI, has launched a tamper-evident provenance logging system for production AI agents, addressing a critical gap in auditing capabilities for AI actions. As AI agents perform tasks such as calling APIs or writing to databases, their actions often lack a clear traceability record, making it challenging to track failures or unauthorized activities. RootSign captures each action in a cryptographic hash chain, ensuring that any tampering is easily detectable. This system requires no changes to existing agent code, broadening its applicability across various environments. Significantly, RootSign offers compliance-grade audit trails that operate with minimal setup. The tool can quickly create a verified audit log in a JSONL format, facilitating integration into CI or automated audits. Upcoming features include a hosted compliance dashboard and enhanced policy enforcement capabilities. The wide-ranging integration options—supporting frameworks like LangGraph, CrewAI, and Model Context Protocol (MCP)—position RootSign as a versatile solution for enhancing accountability within AI workflows. This development not only improves operational transparency but also strengthens the trustworthiness of AI applications in sensitive operations.