CHAP protocol standardizes human-agent collaboration and accountability
Researchers propose CHAP, the Collaborative Human-Agent Protocol, a structured standard for multi-human, multi-agent workflows that turns human overrides, approvals, and handoffs into auditable, signed, replayable events.
Score breakdown
Teams deploying agents in high-stakes domains (claims, code, contracts, clinical decisions) gain a concrete protocol for capturing human oversight as structured, auditable, and legally replayable records rather than ephemeral chat messages.
- 01CHAP is proposed by Arsalan Shahid, Gordon Suttie, and Philip Black as a protocol for accountable human-agent collaboration.
- 02Existing standards MCP (agent-tool access) and A2A (agent-to-agent interoperability) do not define a shared human-agent workspace.
- 03The protocol's Core defines workspaces, participants, tasks, artefacts, and an append-only evidence log.
Arsalan Shahid, Gordon Suttie, and Philip Black present CHAP, the Collaborative Human-Agent Protocol, in response to a gap left by existing standards. While MCP standardizes agent access to tools and data, and A2A standardizes agent-to-agent interoperability, neither defines the shared workspace in which humans and agents perform accountable work together. The paper argues that in current practice, the moment of human judgement — such as when a human edits an agent's draft before it ships — is the most valuable signal in the system, yet it is typically lost in chat threads, ticket comments, or tribal memory.
CHAP addresses this through a small Core consisting of workspaces, participants, tasks, artefacts, and an append-only evidence log.
CHAP addresses this through a small Core consisting of workspaces, participants, tasks, artefacts, and an append-only evidence log. On top of this Core, composable profiles add capabilities including review, modes, routing, deliberation, handoff, identity, signatures, and transparency-backed audit, allowing deployments to adopt only what they need. Under the protocol, an override becomes a structured event carrying a diff, a rationale, and a content hash; a shift handoff becomes a portable envelope rather than a pinned message; and a human approval of an agent's draft becomes a non-repudiable signed decision that can be replayed years later.
The authors have published a specification, reference implementation, conformance suite, and worked examples at the project's GitHub repository, positioning CHAP as a complement to MCP and A2A for production multi-agent systems that cross teams, time zones, and trust boundaries.
Key facts
- 01CHAP is proposed by Arsalan Shahid, Gordon Suttie, and Philip Black as a protocol for accountable human-agent collaboration.
- 02Existing standards MCP (agent-tool access) and A2A (agent-to-agent interoperability) do not define a shared human-agent workspace.
- 03The protocol's Core defines workspaces, participants, tasks, artefacts, and an append-only evidence log.
- 04Composable profiles extend the Core with review, routing, deliberation, handoff, identity, signatures, and transparency-backed audit.
- 05Human overrides become structured events carrying a diff, a rationale, and a content hash.
- 06Human approvals of agent drafts become non-repudiable signed decisions replayable years later.
- 07A specification, reference implementation, conformance suite, and worked examples are available on GitHub.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 9, 2026 · 09:19 UTC. How this works →