Run IDs and Call IDs

HumanLayer uses two types of identifiers to track requests:

​

Run IDs

A run ID identifies a single execution of an agent or workflow. It groups together all function calls and human contacts made during that execution.

hl = HumanLayer(
    run_id="linkedin-inbox-assistant",  # Optional - defaults to auto-generated ID
)

If not specified, HumanLayer will generate a run ID using the agent name or a random identifier.

​

Call IDs

Each individual function call or human contact gets a unique call ID. This ID is used to:

• Track the status of the request
• Fetch updates about approval/response
• Match responses back to the original request

Call IDs are automatically generated but can be specified manually if needed:

# Auto-generated call ID
call = hl.create_function_call(
    spec=FunctionCallSpec(
        fn="send_email",
        kwargs={"to": "user@example.com"},
    ),
)

# Manual call ID (e.g. to match external system IDs like an OpenAI tool_call_id)
call = hl.create_function_call(
    spec=FunctionCallSpec(
        fn="send_email",
        kwargs={"to": "user@example.com"},
    ),
    call_id="email-123",
)

​

Usage in Approvals

When using require_approval or human_as_tool, the IDs are used to track the request through the approval process:

@hl.require_approval()
def send_email(to: str, subject: str, body: str) -> str:
    """Send an email to a customer"""
    # Function call will be tracked with auto-generated run_id and call_id
    return f"Email sent to {to}"

You can view these IDs in the HumanLayer dashboard to track and audit agent activity.