Run Autonomous • AI Agent SDK

run_autonomous: Run an autonomous task execution loop.

run_autonomous

Method

This is a method of the Agent class in the agent module.

Run an autonomous task execution loop.

This method executes a task autonomously, using the agent’s tools and capabilities to complete the task. It handles:

• Progressive escalation based on task complexity
• Doom loop detection and recovery
• Iteration limits and timeouts
• Completion detection (keyword-based or promise-based)
• Optional context clearing between iterations

Signature

def run_autonomous(prompt: str, max_iterations: Optional[int], timeout_seconds: Optional[float], completion_promise: Optional[str], clear_context: bool) -> Any

Parameters

prompt · str · required

The task to execute

max_iterations · Optional

Override max iterations (default from config)

timeout_seconds · Optional

Timeout in seconds (default: no timeout)

completion_promise · Optional

Optional string that signals completion when wrapped in <promise>TEXT</promise> tags in the response

clear_context · bool · default: "False"

Whether to clear chat history between iterations (forces agent to rely on external state like files)

Returns

Returns · Any

AutonomyResult with success status, output, and metadata

Exceptions

ValueError

If autonomy is not enabled

Usage

agent = Agent(instructions="...", autonomy=True)
    result = agent.run_autonomous(
        "Refactor the auth module",
        completion_promise="DONE",
        clear_context=True
    )
    if result.success:
        print(result.output)

Uses

• ValueError
• time_module.time
• isoformat
• datetime.now
• get_recommended_stage
• AutonomyResult
• chat
• clear_history

Source

View on GitHub

praisonaiagents/agent/agent.py at line 1904