MACP Examples
Status: Non-normative (explanatory). In case of conflict, RFC-MACP-0001 is authoritative.
This repository includes illustrative transcripts for each standards-track mode and for policy-governed sessions.
Available example transcripts
| File | Mode | What it shows |
|---|---|---|
examples/decision-mode-session.json | macp.mode.decision.v1 | initialization, SessionStart, proposal, evaluations, objection, votes, and terminal Commitment |
examples/proposal-mode-session.json | macp.mode.proposal.v1 | offer, counteroffer, dual acceptance, and bound agreement |
examples/task-mode-session.json | macp.mode.task.v1 | bounded delegation, progress, completion, and bound task outcome |
examples/handoff-mode-session.json | macp.mode.handoff.v1 | responsibility transfer with context handoff and acceptance |
examples/quorum-mode-session.json | macp.mode.quorum.v1 | threshold approval leading to a final commitment |
examples/policy-decision-session.json | macp.mode.decision.v1 | policy-governed Decision Mode with supermajority voting, quorum, and critical-severity veto (RFC-MACP-0012) |
examples/policy-registration-exchange.json | N/A (gRPC exchange) | dynamic policy registration request and response (RFC-MACP-0012 Section 7) |
Discovery, lifecycle, and runtime examples
Single-document examples that exercise non-transcript schemas:
| File | Schema | What it shows |
|---|---|---|
examples/discovery/agent_manifest.json | macp-agent-manifest.schema.json | agent identity, supported modes, transport endpoints |
examples/discovery/mode_descriptor.json | macp-mode-descriptor.schema.json | mode advertisement with message types and schema URIs |
examples/discovery/policy_descriptor.json | macp-policy-descriptor.schema.json | governance policy advertisement (RFC-MACP-0012) |
examples/discovery/session_metadata.json | macp-session-metadata.schema.json | SessionMetadata returned by GetSession |
examples/discovery/session_lifecycle_event.json | macp-session-lifecycle-event.schema.json | SessionLifecycleEvent emitted by WatchSessions |
examples/discovery/run_descriptor.json | macp-run-descriptor.schema.json | scenario-agnostic run descriptor for a control-plane POST /runs |
examples/discovery/agent_bootstrap.json | macp-agent-bootstrap.schema.json | bootstrap payload written to MACP_BOOTSTRAP_FILE before the initiator agent starts |
Example shape
Each transcript is an ordered JSON document with:
- top-level metadata such as
description,protocol_version,mode, andsession_id, - an ordered
messagesarray containing full Envelope objects in replay order, - envelopes and initialization payloads written in replay order.
The ordering of the messages array is the replay order.
Note: The policy-governed examples (
policy-decision-session.json,policy-registration-exchange.json) use a simplifiedtranscriptarray with sequence numbers (seq) and condensed message representations. This format is intended to illustrate governance evaluation flow rather than serve as a wire-format reference. For canonical Envelope structure, refer to the standard mode transcripts.
What the examples demonstrate
Across the transcripts, the examples demonstrate:
- explicit initialization and capability negotiation,
- explicit
SessionStart, - session-scoped mode messages,
- version binding suitable for replay,
- a terminal
Commitmentthat resolves the Session, - policy resolution, governance evaluation, and deterministic commitment decisions (RFC-MACP-0012).
The important property is not the specific business scenario in each example. It is that every bound coordination event exists as a bounded transcript with a start, a lifecycle, and a terminal message.