Agent Delegation

Delegation tools let agents hand off work to other agents. There are two modes: fire-and-forget with delegate_task, and synchronous with run_subtask.

delegate_task

Delegates a task to another agent. The calling agent does not wait for the result. When the target agent finishes, the result is posted directly to the chat.

delegate_task(
  target_agent: "researcher",
  title: "Research market trends",
  instruction: "Find recent data on AI adoption in healthcare"
)

Parameters

Parameter	Type	Required	Description
target_agent	string	yes	Name of the agent to delegate to
title	string	yes	Short title for the task
instruction	string	yes	Detailed instructions for the target agent
run_at	string	no	ISO 8601 datetime to defer execution

run_subtask

Runs a task on another agent and waits for the result. The calling agent’s tool loop pauses until the target agent completes the work, then the result is returned to the calling agent so it can continue.

run_subtask(
  target_agent: "developer",
  title: "Generate API client",
  instruction: "Create a Python client for the payments API based on the OpenAPI spec"
)

Parameters

Parameter	Type	Required	Description
target_agent	string	yes	Name of the agent to delegate to
title	string	yes	Short title for the task
instruction	string	yes	Detailed instructions for the target agent
run_at	string	no	ISO 8601 datetime to defer execution

delegate_task vs. run_subtask

	delegate_task	run_subtask
Calling agent waits	No	Yes
Result goes to	Chat directly	Calling agent
Use when	You want to hand off and move on	You need the result to continue

read_skill

Loads the full content of a pre-defined skill by name. Skills are prompt and tool configurations that extend what an agent can do without delegating to a separate agent.

read_skill(name: "code-review")

Parameters

Parameter	Type	Required	Description
name	string	yes	The name of the skill to load

Choosing delegation targets

The platform provides each agent with a list of available agents. The agent uses the name and description of each available agent to decide who to delegate to.

This is why writing clear agent descriptions matters. They help other agents make good delegation decisions.

Related

• Agent Delegation: architectural overview
• Agent Types: built-in agents available for delegation