Repository avatar
AI Tools
v0.3.5
active

chap

io.github.BrightbeamAI/chap

Auditable records of human decisions over AI agent work. Approvals, edits, overrides, escalations.

Documentation

Collaborative Human-Agent Protocol (CHAP)

Latest release PyPI package npm package CI status Specification licensed CC BY 4.0 Code licensed Apache 2.0

The open protocol for humans and agents doing accountable work together.

CHAP gives approvals, overrides, handoffs and escalations one shared, auditable shape across MCP and A2A.

Start here · Concepts · Install · 90-second tour · MCP quickstart · Scenarios · Implementations · Wiki · Discussions · Website · Paper

Star CHAP on GitHub to help more implementers find and test the protocol


Same scenario, two stacks. Without CHAP: six tools holding fragments of one decision (OpenAI logs expired, Zendesk thread, Slack scrolled past, Linear comments, webhook tail, Notion runbook), 45 minutes across four UIs to answer 'what did the agent draft and why did we approve it?'. With CHAP and its hash chain switched on: three hash-linked envelopes (task.create → artefact → decide.override) joined by prev_hash, one audit.read call, 30 seconds.


Work is moving towards teams in which hundreds of agents and hundreds of people take part. Agents draft, triage and recommend; people approve, correct, overrule and escalate. Those human decisions make the work accountable, yet they are usually recorded only in chat threads, ticket comments and memory.

CHAP gives those decisions a defined structure and an auditable record: what the agent produced, what the person decided and why, in one log that can be queried and verified.

  • Structured overrides. When a person edits an agent's work, the edit is recorded with a diff, a rationale and tags your team defines, so corrections can be analysed later.
  • A tamper-evident log. Every change is recorded in order. With the hash chain on, an altered entry is detected on verification, provided a recent chain head is kept out of the operator's reach (SECURITY.md §5). The history survives key rotation, expired vendor logs and staff changes.
  • Signatures and identity. Optional extensions, called profiles, require every call to be signed, bind signing keys to verified identities, and submit entries to a transparency log.
  • Alongside MCP and A2A. MCP connects agents to tools and A2A connects agents to each other; CHAP records the work agents share with people.

Concepts

CHAP is a protocol: a set of calls that agents, people and their tools send to a coordinator, the service that mediates a workspace. The coordinator checks each call against the workspace's rules, answers it, and records each accepted change in the workspace's audit log. The TypeScript and Python packages below are coordinators, to embed in your own process or run as a server.

TermWhat it is
WorkspaceA named context for one body of work: its members, its tasks, and the profiles it uses
ParticipantA member, named by a URI such as human:alice@example.org or agent:triage; group: names a set of members
TaskA unit of work, assigned to a participant
ArtefactWhat a participant produces in a task: a draft, a decision, an override
EnvelopeOne call, as a JSON-RPC 2.0 message; its method is the verb, such as task.create or decide.approve
Audit logThe ordered record of every change to the workspace; with the hash chain on, each entry carries the hash of the one before it

The verbs read as their names. A task is created, completed and sent for review (task.create, task.complete, review.request). A reviewer approves, rejects or overrides it (decide.approve, decide.reject, decide.override). Work is handed off or escalated (handoff.propose, escalate.raise), and audit.read returns the record.

Core and profiles. Every coordinator implements Core (core/1.0): workspaces, participants, tasks and the audit log. Everything else is an optional profile, named with its version. review/1.0 adds review requests and decisions. Others add quick questions to a person (whisper/1.0), group votes (deliberation/1.0), handoffs (handoff/1.0), pausing and rollback (control/1.0), routing (routing/1.0), shadow and trial modes (modes/1.0), signed calls (security-signed/1.0), verified identities (identity-oidc/1.0, identity-vc/1.0) and a transparency log (audit-scitt/1.0). A workspace advertises the profiles it uses, and the coordinator refuses calls that belong to any other profile. core/SPEC.md fits Core on one page, profiles/PROFILES.md lists the profiles, and GLOSSARY.md defines every term.

Install

ForRun
TypeScriptnpm install @brightbeamai/chap-coordinator
Pythonpip install chap-coordinator
A new project with a review desknpx create-chap-app my-gate

Each package implements Core and every profile; a new workspace advertises core/1.0 and review/1.0 unless you name others. START_HERE.md is the full guide: a first recorded decision from a clone of this repository, with Python alone, and every other way in.

The override envelope

The override envelope records a reviewer's change to an agent's output. Each field is annotated below:

Anatomy of a decide.override envelope, with each field annotated: task_id links to the review chain, from carries queryable identity, logical_id survives revision, intent_preserved separates refining from substituting overrides, diff is RFC 6902 JSON Patch, rationale is the 'why' alongside the 'what', tags are structured supervision data.

intent_preserved separates a refining override, which keeps the agent's decision and rewrites its wording, from a substituting one, which decides differently. tags are the small vocabulary a team agrees on and counts by later. ABOUT.md says what each points at.

Status

CHAP 0.3 is a public draft: a small Core and optional profiles (SPECIFICATION.md), with reference coordinators in TypeScript and Python that a differential fuzzer checks answer and log alike. Before 1.0 a minor release may break things, and its changelog lists each break with a migration. ABOUT.md says what is built so far, and ROADMAP.md sets out the milestones to 1.0.

Read this next

ReadFor
START_HERE.mdA first recorded decision, and every way in
docs/tour.mdThe 90-second tour: a clip of one review, and the code for each step in TypeScript and Python
IN_PRACTICE.mdScenarios, from a solo developer with Cursor to GMP-regulated manufacturing
ABOUT.mdHow Core and the profiles are adopted, what the override envelope's fields point at, how CHAP sits with MCP and A2A, the standards it reuses, the status and the repository's layout
SPECIFICATION.mdCore and every profile, in one document
core/SPEC.mdCore, on one page
profiles/PROFILES.mdThe optional profiles, and how they combine
GLOSSARY.mdEvery term
IMPLEMENTATIONS.mdThe packages, the framework bridges and other implementations
packages/chap-analytics/The audit log as pandas tables and charts
SECURITY.mdThe threat model, and how to report a vulnerability
CHANGELOG.mdWhat changed in each release
CONTRIBUTING.mdHow to contribute; GOVERNANCE.md for how the protocol evolves
The technical reportThe architecture, profile semantics and threat model, with the scenarios as JSON traces

Cite

If you reference CHAP in academic or technical work, please cite the technical report.

BibTeX
@techreport{chap2026,
  author      = {Shahid, Arsalan and Suttie, Gordon and Black, Philip},
  title       = {Collaborative Human-Agent Protocol (CHAP): An open protocol for auditable, structured multi-human and multi-agent collaboration},
  institution = {Brightbeam AI},
  year        = {2026},
  type        = {Technical Report},
  number      = {arXiv:2606.09751},
  url         = {https://arxiv.org/abs/2606.09751}
}

CC BY 4.0 (specification text, see LICENSE-SPEC.md) · Apache 2.0 (everything else) · Any language, any deployment.