CHAP Protocol Aims to Make Human Edits in AI Agent Workflows Auditable

A new open protocol called CHAP (Collaborative Human-Agent Protocol) has been introduced to address a critical gap in AI accountability: human edits and decisions made during AI agent workflows are rarely recorded in a durable, structured way. Developed by BrightbeamAI, CHAP treats human judgements — such as overrides, approvals, and shift handoffs — as structured, traceable events stored in an append-only evidence log. The protocol is designed to complement existing standards like MCP and A2A, which govern tool access and agent-to-agent communication but do not cover shared human-agent workspaces. With the EU AI Act's oversight requirements taking effect, organizations deploying AI in regulated domains such as insurance, healthcare, and legal services will need verifiable records of human decision-making. The specification is publicly available under CC-BY 4.0, with Apache 2.0-licensed code and adapters for major agent frameworks including Pydantic AI, LlamaIndex, AG2, and Google ADK.