Local control plane¶
AIT's local control plane coordinates explicit connections, live interception, controlled labs, traffic records, and the Cockpit. It does not replace the data plane: Seam owns protocol capture and delivery, Assay owns oracle adjudication, and meshmapper owns deterministic topology hypotheses.
Start with readiness¶
doctor checks the source checkout and packaged peer tools. It does not contact
a target.
Run the control plane¶
The API and Cockpit bind loopback. A random bearer token protects the local surface and is stored under the private workspace state. Moving the data-plane listener to another interface does not move the control plane.
Create one explicit path¶
ait connections add \
--name specialist \
--protocol a2a_jsonrpc \
--upstream http://127.0.0.1:9000
ait connections test CONNECTION_ID
ait intercept start --connection CONNECTION_ID
The start command prints the listener address to give the sender. AIT does not discover a route and silently insert itself.
Operate a bounded exchange¶
The pending queue contains complete decoded messages. The immutable sender representation, operator decision, delivered representation, correlation data, and response remain distinct in the record.
Use the controlled lab¶
ait lab start --exercise delegated-a2a-message --no-open
ait lab trigger
ait lab status
ait lab reset
ait lab stop
The lab is the safest place to learn the lifecycle. It uses real protocol adapters and separate processes, but its roles, policy, and effect are authored fixtures. It is not external validation.
Export and hand off¶
ait intercept export SESSION_ID \
--format jsonl \
--redacted \
--out session.jsonl
ait traffic tail --run RUN_ID --limit 50
Redacted export omits bodies and replaces credential values. Raw transcripts remain sensitive and should be retained only when the assessment requires them.
Reach the peer tools¶
The advanced namespace exposes reviewed peer-tool operations without pretending their evidence is interchangeable:
- Seam can verify and trace what crossed the boundary.
- meshmapper can turn records into unvalidated path hypotheses.
- Assay can compare a baseline and observation against an explicit oracle.
No component may promote its own output into a stronger evidence tier merely because the output exists. See System architecture and R/R/R methodology for those boundaries.