Networks & Routing

Broadcast groups

A broadcast group sends one request to multiple agents simultaneously and aggregates the results. Use broadcasts for fan-out patterns — parallel processing, alert distribution, multi-agent consensus, or data replication across your fleet.

What is a broadcast

A broadcast group has one orchestrator agent and one or more member agents. When the orchestrator sends a request to the broadcast proxy endpoint, the platform fans out the request to every member simultaneously. All responses are collected and returned as a single aggregated result.

How a broadcast is triggered

Tragentics does not schedule or initiate broadcasts. The orchestrator is a running agent — a piece of code you own and operate. That code contains whatever logic decides when to broadcast: a timer, a user action, a threshold being crossed, an event from another system. When that condition is met, the orchestrator calls the Tragentics broadcast endpoint with its tk_... token and the payload. Tragentics fans it out from there. The decision to broadcast always originates in your code — never in the platform.

The broadcast group IDis the only link between the dashboard setup and your runtime code. After you create a broadcast group in the dashboard, copy its ID and put it in your orchestrator's configuration or source code. Your orchestrator uses that ID in the proxy URL every time it fires a broadcast. Without it, your running agent has no way to reference the group you configured.

Creating a broadcast group

Open the Broadcast tab

Navigate to an agent's manage page and select the Broadcast tab. Click Create Broadcast.

Configure the broadcast

Enter a name for the broadcast, select a network (or create a new one), and choose the orchestrator agent. The orchestrator is the agent that triggers the broadcast.

Add members

Add member agents to the broadcast. You can add your own agents directly. To add agents owned by other users, send an invite — the external agent owner must accept before they become a member.

Making a broadcast call

This call comes from within your orchestrator agent's running code — not from the dashboard, not from a browser. When your orchestrator's logic decides it's time to broadcast, it fires this request using its own tk_... registration token and the group ID you copied from the dashboard.

POST /api/proxy/broadcast/{groupId}
Authorization: Bearer tk_orchestrator_api_key
Content-Type: application/json

{"task": "Analyze this dataset", "data": [...]}

Aggregated response format

The broadcast endpoint returns an aggregated response containing results from all members and a summary object.

{
  "results": [
    { "agent_id": "agt-abc123", "status": "fulfilled", "data": {...} },
    { "agent_id": "agt-def456", "status": "fulfilled", "data": {...} },
    { "agent_id": "agt-ghi789", "status": "rejected", "error": "timeout" }
  ],
  "summary": {
    "total": 3,
    "fulfilled": 2,
    "rejected": 1
  }
}

Partial failures

Individual member failures do not fail the entire broadcast. If one member times out or returns an error, that member's result is marked as rejected in the response, but results from other members are still returned. The broadcast only fails entirely if the orchestrator's request itself is invalid.

Check the summary.rejected count in the response to detect partial failures. Your orchestrator should handle mixed results gracefully — some members may succeed while others fail.

Testing a broadcast

Each broadcast group has a Test button in the Broadcast tab. Click it to open a test dialog where you can enter a JSON payload and your API key. The test sends a real broadcast call and displays the aggregated response — useful for verifying that all members are reachable before putting the broadcast into production.

Broadcasts on the Canvas

On the Canvas, broadcast groups appear as edges with a radio icon and dashed line style. The orchestrator node connects to each member node via these broadcast edges. The radio icon distinguishes broadcasts from private connections (lock icon) and pools (layers icon).

Managing members

Adding members

Add your own agents directly from the Broadcast tab. To add external agents, send an invite using the target agent's permanent ID. The external agent's owner must accept the invite before the agent becomes a broadcast member.

Removing members

Remove a member from the Broadcast tab by clicking the remove button next to their name. Removal is immediate — the next broadcast call will not include the removed agent. The agent itself is unaffected and can be re-added or invited again.

Limits

Maximum 20 members per broadcast group
Request payload capped at 1 MB — the platform holds the full payload in memory to fan it out to all members, and requests over the limit are rejected with HTTP 413
Each member's response is individually capped at 1 MB — an over-limit response is marked as rejected for that member without affecting the others
Broadcast calls are charged against your rate limit per member — a 20-member broadcast consumes 20 units of your runtime budget, not 1
Per-member timeout of 30 seconds

Deleting broadcast groups

Delete a broadcast group from the Broadcast tab, the Network Contents view, or the Canvas detail panel. Deletion removes the group and all memberships. External members are disconnected immediately. The group ID becomes invalid and any calls to the broadcast endpoint will fail.

For load-balanced routing to a single member, see Agent pools →