Skip to content

Orchestration & Planning Tools

The orchestration toolkit provides a plan + task graph (dependencies) that helps an agent coordinate multi-step work.

It is designed to be:

  • Deterministic and auditable (no hidden planning logic)
  • Dependency-aware (depends_on)
  • Tool-call friendly (returns JSON strings)

Auto-loaded when using any toolkit

enhancedtoolkits auto-loads the orchestration singleton and injects orchestration functions into every toolkit via StrictToolkit.

That means:

  • If you add any toolkit (e.g. ReasoningTools) to your agent, you also get these orchestration tools.
  • Injected tool names are prefixed to avoid collisions:
  • orchestrator_create_plan()
  • orchestrator_add_task()
  • orchestrator_update_task_status()
  • orchestrator_next_actions()
  • orchestrator_summarize_progress()
  • orchestrator_reset_plan()

Optional: add the toolkit directly

You can also add OrchestrationTools directly to your agent. In that case, the tool names are the “clean” names:

  • create_plan()
  • add_task()
  • update_task_status()
  • next_actions()
  • summarize_progress()
  • reset_plan()

🤖 AI Agent Setup (Agno)

from agno.agent import Agent
from enhancedtoolkits import ReasoningTools

# Orchestration tools are auto-injected into every toolkit instance.
agent = Agent(
    name="Planner",
    model="gpt-4",
    tools=[ReasoningTools()],
)

✅ Example workflow

import json

class Session:
    pass

agent = Session()

# create a plan
plan_json = ReasoningTools().orchestrator.create_plan(
    agent_or_team=agent,
    goal="Write docs + run validation",
    tasks=[
        {"title": "Update docs", "priority": 5},
        {"title": "Run link checks", "priority": 3, "depends_on": [1]},
    ],
)
print(json.loads(plan_json)["plan"]["plan_id"])

# ask for next task(s)
actions_json = ReasoningTools().orchestrator.next_actions(agent_or_team=agent, max_actions=1)
print(json.loads(actions_json)["actions"])

# mark done
ReasoningTools().orchestrator.update_task_status(agent_or_team=agent, task_id=1, status="done")

# summarize
print(ReasoningTools().orchestrator.summarize_progress(agent_or_team=agent))

API Reference