Simple Orchestrator

A minimal orchestrator example that demonstrates the basic pattern for coordinating sub-agents. This is a good starting point if you want to understand how orchestrators work before building something more complex like the News Orchestrator.

Prerequisites

Important: You must set up at least two sub-agents before creating this orchestrator. Each sub-agent must be a fully configured agent with a running server. The orchestrator cannot function until its sub-agents are available.

This example works with any sub-agents you have available. You just need at least two agents that the orchestrator can delegate to. For example, you could use two of the news agents from the agents examples, or any other agents that you have set up.

Setting Up the Orchestrator

  1. On the Agents pane, click Create an Agent Orchestrator.
  2. Create a new experiment. Everything can be left as the default settings except the Agents tab.
  3. Agents: Add any sub-agents that you would like (at least two are recommended).
  4. Click Create.
  5. Start the orchestrator’s server.

How It Works

The no-code orchestrator uses RAM’s built-in orchestrator agent, which does the following:

  1. Receives a user query
  2. Uses the LLM to decide which sub-agents to query based on their agent cards
  3. Sends queries to the selected sub-agents via A2A (Agent-to-Agent communication)
  4. Synthesizes the sub-agent responses into a single answer

This is the simplest way to set up an orchestrator in RAM. For more control over the orchestration logic (such as custom routing, multi-step research, and specialized prompts), see the News Orchestrator example that uses a code template.

Back to top

Open source under the Apache License 2.0. Provided AS IS, without warranties, responsibility, or support from SAS except where applicable law requires otherwise.