Halo: Open-Source Tamper-Evident Runtime Evidence for AI Agents

Halo provides a verifiable audit trail for AI agent actions

Halo is an open-source framework designed to create tamper-evident runtime records for AI agents. It transforms every agent action—including tool calls, model calls, data access, and approvals—into an append-only, hash-chained log. This allows any party to verify that the agent's operational history has not been altered, removing the need to trust the vendor running the agent.