Jarvis ARM
The First ARM.

Jarvis identifies agent-originated traffic at the point of entry by protocol type. MCP, A2A, and UCP sessions are classified and separated from human sessions before routing begins. Every agent interaction is tagged at ingestion so the system always knows what it is handling.

The dealer's own transacting agent is in production. When a buyer's agent enters the pipeline, the ARM deploys the dealer's pre-configured agent to meet it autonomously within set parameters. The A2A protocol is the channel through which the dealer's agent and the buyer's agent negotiate directly, machine to machine. Agent-to-agent traffic is live in the audit log.

Pricing floors, doc fees, registration fees, and tax disclosure are encoded into the protocol schema and managed per rooftop by the dealership. The ARM holds these parameters through every quote request. The get_pricing_disclosure tool and the request_quote out-the-door estimate consume the saved fees on every call. No concessions outside the dealer's defined boundaries without human authorization. UCP carries the commercial transaction envelope that enforces the terms.

Agent leads and human leads travel separate paths from the moment they enter the pipeline. The protocol classifier in Jarvis applies distinct logic to each traffic type. Human leads route to the CRM. Agent leads route to the ARM pipeline. Same dealership. Same inventory. Two parallel systems.

Independent reporting on all agent activity runs as a separate line from human analytics. Requests, error rate, active agents, latency, throughput by protocol, tool mix, and audit log are all live in the Jarvis dashboard. AG-UI is the bi-directional protocol that connects the agent backend to the human-facing interface in real time, streaming agent activity to the operator and carrying operator inputs back to the agent. A2UI is Google's declarative UI specification that defines the structure of the interface components the agent returns. Together they solve the human-in-the-loop problem: machine-speed agent activity becomes a readable, actionable operator dashboard.

The knowledge base update loop allows the dealer's agent to refine its response logic based on what worked and what did not across previous negotiations. When a buyer's agent applies a pattern the dealer's agent has seen before, it recognizes it and adjusts. The audit trail being built now is the foundation for this layer. AG-UI carries the negotiation event stream that will feed the pattern recognition engine as it comes online.

Agent identity is verified cryptographically via Auth0 M2M authentication before any tool access is granted. Named agents are checked against trusted registries. Legitimate consumer advocacy agents, scraper bots, and malicious probing agents each receive a distinct response protocol. No identity, no transaction.

Jarvis ARM operates a separate agentic lane that runs alongside the dealership stack, not through it. Agent-to-agent transactions execute end to end at machine speed without a human in the loop. The protocol handles discovery, quote, reservation, and deal structure autonomously, drawing on tools in the dealership stack when needed. AG-UI and A2UI surface the live activity in the operator's existing environment, so any human who wants to watch or act has full visibility. Discovery and delivery stay with humans because that is where the emotion lives. The friction in the middle is removed by design.

Every agent interaction is recorded in the audit log with a cryptographic trace. Agent identity, protocol, tool called, status, latency, and trace ID. Persistent. Non-repudiable. Queryable. Every signed UCP and A2A envelope creates a permanent record. The ARM never starts from zero. When a buyer's agent returns, the system knows it.

Forward projection of agent traffic volume and gross margin impact based on accumulated interaction data. The audit trail and throughput data being built now are the foundation. As the projection layer comes online, AG-UI will carry forecast data to the frontend in real time using A2UI-structured components, giving leadership a live view of what the agent pipeline will produce next quarter.