Experimental Features

Inter-Agent Communication

Agents can discover and communicate with other agents in your Fleet. This enables specialized agents that each own a domain of knowledge.

How It Works

Agents have access to two tools:

• agent_list — Get a list of all agents with their IDs, names, and status
• agent_message — Send a message to another agent by ID

When an agent receives a message, it wakes up, processes the request, and can respond back.

Example: Domain Experts

Create agents that specialize in different areas:

"Finance Agent" — Knows your budget, expenses, financial goals

"Projects Agent" — Tracks your projects, deadlines, status

"Contacts Agent" — Manages contact information and relationships

When you ask a general agent a question like "What's the status of the Johnson project and do we have budget for it?", it can:

1. Message the Projects Agent to get project status
2. Message the Finance Agent to check budget
3. Combine the responses into a single answer

Setting It Up

1. Create specialized agents with clear domains
2. Give each agent context about what it knows
3. Create a "coordinator" agent that can delegate to specialists
4. Ask the coordinator — it routes questions to the right experts

Benefits

• Separation of concerns — Each agent focuses on one area
• Persistent context — Domain agents maintain their own knowledge
• Parallel work — Multiple agents can work simultaneously
• Scalable — Add new domain experts as needed

Note

Experimental features may change or be removed in future versions. Use them for exploration, not critical workflows.