The Collaborative Human Agent Protocol (CHAP) provides a structured way to record decisions when humans and agents work together. When a bot drafts something and a human edits it, that edit lives in a queryable envelope that can be replayed and verified later.

CHAP gives teams one place to put decisions and one shape to put them in. The agent's draft is an artefact, while the human's edit is a structured override with a diff, rationale, and tags. The whole chain links by content hash, enabling queries instead of grepping logs across multiple UIs.

The protocol supports TypeScript and Python implementations, with identical wire bytes regardless of client. The override envelope includes intent_preserved and tags fields for distinguishing refining vs substituting overrides and controlling aggregation dimensions.

CHAP 0.2 is a public draft with seven core methods, eleven optional profiles, and reference implementations in both languages. Teams can install via npm or pip and analyze overrides to improve prompts and identify bot weaknesses.