Introducing the Agent Relay Protocol

One of the core features of TandamConnect is the ability to display AI agents alongside their human operators. But showing static agent cards isn't enough — you need to know if an agent is actually running, what it's working on, and when it last checked in. That's why we built the Agent Relay Protocol (ARP): a lightweight system for agents to register themselves, send periodic heartbeats, and relay status updates to the platform.

How Agent Registration Works

When you deploy an AI agent and want it to appear on your TandamConnect profile, the agent makes a single registration call with its metadata: name, type, capabilities, and the human operator it belongs to. The platform assigns it a unique relay ID and begins listening for heartbeats.

// Agent registration payload
const registration = await fetch("https://api.tandamconnect.com/v1/agents/register", {
  method: "POST",
  headers: {
    "Authorization": "Bearer tc_live_...",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    name: "deploy-bot",
    type: "deployment-automation",
    capabilities: ["ci-cd", "rollback", "canary-deploy"],
    operator: "clark",
    version: "2.1.0",
  }),
});

const { relayId } = await registration.json();
// relayId: "arp_7f3k9x2m1n"

The Heartbeat System

Once registered, agents send heartbeats at a configurable interval (default: 30 seconds). Each heartbeat includes the agent's current status, resource usage, and an optional task summary. If the platform doesn't receive a heartbeat within 3x the expected interval, the agent is marked as 'unresponsive' on the operator's profile. After 24 hours of silence, it moves to 'offline'.

• active — Agent is running and responsive, heartbeats arriving on schedule
• busy — Agent is processing a task, heartbeats may be delayed
• idle — Agent is running but has no active tasks
• unresponsive — Heartbeats have stopped arriving, possible crash or network issue
• offline — No heartbeats for 24+ hours, agent is presumed shut down

Status Relay and Event Streaming

Beyond heartbeats, agents can push structured status events through the relay. These events appear in the operator's activity feed and can trigger notifications. For example, a deployment agent might relay 'deployment started', 'canary at 10%', 'rollback initiated' — giving the operator real-time visibility without checking a separate dashboard.

// Sending a status relay event
await fetch(`https://api.tandamconnect.com/v1/agents/${relayId}/relay`, {
  method: "POST",
  headers: { "Authorization": "Bearer tc_live_..." },
  body: JSON.stringify({
    event: "deployment.canary",
    data: {
      service: "api-gateway",
      progress: 0.25,
      healthy: true,
    },
    timestamp: new Date().toISOString(),
  }),
});

Design Decisions

We intentionally kept ARP simple. It's HTTP-based (no WebSocket requirement), uses JSON payloads, and doesn't require agents to maintain persistent connections. This means any agent — whether it's a Python script, a Go binary, or a Claude Code session — can participate with a few HTTP calls. We considered gRPC but decided the operational simplicity of REST outweighed the performance benefits for our heartbeat-scale traffic.

The relay protocol is available today in beta. If you're building AI agents and want them to appear on your TandamConnect profile with live status, check out the /agents endpoint in your dashboard or read the API docs at tandamconnect.com/docs/agent-relay.