Skip to main content

​
What Is It?

The Agent Communication toolkit lets your agent talk to other agents in your project (or across projects) so they can work together. It has three tools: Message Agent, Delegate Task, and Agent Status.

​
Message Agent

​
When to Use

Use the Message Agent when your agent needs input, answers, or expertise from another agent to complete its task. For example:
  • A customer support agent needs a product specialist agent to answer a technical question.
  • A reporting agent needs a data analysis agent to pull and summarize numbers from a database.
  • A general-purpose agent needs a domain expert agent to verify or enrich its response.

​
How to Use

  • In the Agent Builder, add the Agent Communication toolkit to your agent.
  • Click Add Agent Communication and select the target agent you want to communicate with.
  • Write a clear description that tells your agent when to reach out, e.g., “Use this agent when the user asks about product specifications or compatibility.”
  • Enable the Message Agent tool in the tool selector.
At runtime, your agent will automatically send a message to the target agent when the situation matches the description. The target agent processes the message using its own tools and knowledge, and sends a response back. Your agent then uses that response to continue the conversation.

​
Delegate Task

​
When to Use

Use Delegate Task when your agent receives a request that is better handled entirely by another specialized agent. Instead of trying to answer itself, the agent hands off the task. For example:
  • A front-desk agent receives a request to generate an invoice - it delegates the task to a billing agent.
  • A manager agent receives a data cleanup request - it delegates to a data processing agent.
  • A triage agent categorizes incoming requests and delegates each one to the right specialist agent.

​
How to Use

  • Add the Agent Communication toolkit and configure a target agent (same steps as Message Agent above).
  • Enable the Delegate Task tool in the tool selector.
  • In the description, explain what kind of tasks should be delegated to this agent, e.g., “Delegate all invoice generation and billing-related tasks to this agent.”
At runtime, when the agent encounters a task matching the description, it delegates the task to the target agent and returns the result.

​
Agent Status

​
When to Use

Use Agent Status when your agent needs to check on another agent before deciding to communicate or delegate. For example:
  • Before delegating a long-running task, the agent checks whether the target agent is available.
  • In a multi-step workflow, the agent checks the status of a previous step handled by another agent before proceeding.

​
How to Use

  • Add the Agent Communication toolkit and configure a target agent.
  • Enable the Agent Status tool in the tool selector.
At runtime, the agent can check the status of the target agent and decide its next action based on the result.

​
Things to Keep in Mind

  • An agent cannot communicate with itself - the system prevents this to avoid infinite loops.
  • Each target agent is limited to 25 tool calls per message, and the overall limit across all agent communications in a single turn is 100 tool calls.
  • The description you provide is critical - it is what the AI reads to decide when to use the tool. Make it specific and clear.
  • You can enable confirmation policies on any tool if you want the user to approve the action before it runs.