Task Automation Agent with Deep Agents

from deepagents import create_deep_agent
from langchain_core.tools import tool

@tool
def create_task(title: str, description: str, assignee: str = "") -> str:
    """Create a new task in the project."""
    task_id = project.create_task(title=title, description=description, assignee=assignee)
    return f"Task created: #{task_id} - {title}"

@tool
def list_tasks(status: str = "open") -> str:
    """List tasks filtered by status (open, in_progress, done)."""
    tasks = project.list_tasks(status=status)
    return "\n".join(f"#{t.id} [{t.status}] {t.title} -> {t.assignee}" for t in tasks)

@tool
def update_task(task_id: int, status: str) -> str:
    """Update task status."""
    project.update_task(task_id, status=status)
    return f"Task #{task_id} updated to {status}"

agent = create_deep_agent(
    model="anthropic:claude-sonnet-4-5-20250929",
    tools=[create_task, list_tasks, update_task],
    system_prompt="You are a project management assistant. Help organize tasks: create new ones, track progress, and update statuses. Be proactive about suggesting task breakdowns.",
)

result = agent.invoke({
    "messages": [("user", "Break down 'Launch new website' into subtasks and assign them to the team")]
})
print(result["messages"][-1].content)