How to Fix 'async event loop error' in CrewAI (Python)

What the error means

If you’re seeing RuntimeError: This event loop is already running or a CrewAI-flavored variant like async event loop error, you’ve got an asyncio lifecycle problem, not a CrewAI-specific bug. It usually shows up when you call async code from a place that already has an active event loop, like Jupyter, FastAPI, Streamlit, or another async framework.

In practice, this happens when you mix asyncio.run(), await, and sync wrappers in the wrong place. CrewAI agents, tasks, and crews can run fine in Python, but the moment you nest event loops or block inside one, Python throws.

The Most Common Cause

The #1 cause is calling asyncio.run() inside an environment that already has a running loop.

This is common when people copy a working script into Jupyter or a web app and keep the same entrypoint pattern. With CrewAI, the mistake usually looks like wrapping crew.kickoff() or an async tool call the wrong way.

Broken vs fixed

# BROKEN
import asyncio
from crewai import Agent, Task, Crew

agent = Agent(
    role="Researcher",
    goal="Find facts",
    backstory="You research company data."
)

task = Task(
    description="Summarize the latest quarterly report.",
    agent=agent
)

crew = Crew(agents=[agent], tasks=[task])

# This breaks in Jupyter / FastAPI / any running loop
result = asyncio.run(crew.kickoff())
print(result)

# FIXED: async context
from crewai import Agent, Task, Crew

agent = Agent(
    role="Researcher",
    goal="Find facts",
    backstory="You research company data."
)

task = Task(
    description="Summarize the latest quarterly report.",
    agent=agent
)

crew = Crew(agents=[agent], tasks=[task])

async def main():
    result = await crew.kickoff()
    print(result)

# In a script:
# import asyncio
# asyncio.run(main())

If you’re in Jupyter, don’t use asyncio.run() at all. Just do:

result = await crew.kickoff()

That’s the clean fix for most cases.

Other Possible Causes

1. Mixing sync and async tool functions

If your CrewAI agent uses tools that are declared sync but internally call async code, you’ll get loop errors or hanging behavior.

# BROKEN
from crewai_tools import tool
import asyncio

@tool("fetch_data")
def fetch_data():
    return asyncio.run(fetch_remote_data())  # bad inside running loop

# FIXED
from crewai_tools import tool

@tool("fetch_data")
async def fetch_data():
    return await fetch_remote_data()

If the tool framework expects sync functions only, move the async work outside the tool or wrap it at the app boundary, not inside the tool itself.

2. Running CrewAI inside FastAPI request handlers with nested loops

FastAPI endpoints are already async. If you call asyncio.run() there, Python will complain immediately.

# BROKEN
from fastapi import FastAPI
import asyncio

app = FastAPI()

@app.get("/run")
async def run_crew():
    result = asyncio.run(crew.kickoff())
    return {"result": str(result)}

# FIXED
from fastapi import FastAPI

app = FastAPI()

@app.get("/run")
async def run_crew():
    result = await crew.kickoff()
    return {"result": str(result)}

3. Using Streamlit callbacks with blocking async code

Streamlit reruns scripts often and doesn’t like nested event-loop management.

# BROKEN
import streamlit as st
import asyncio

if st.button("Run"):
    result = asyncio.run(crew.kickoff())
    st.write(result)

# FIXED
import streamlit as st

if st.button("Run"):
    result = await crew.kickoff()
    st.write(result)

If your Streamlit version or runtime doesn’t support top-level await, move the async work into a separate service layer and call it through a sync wrapper outside Streamlit’s render path.

4. Old Python or incompatible dependency versions

Some event-loop issues come from mismatched versions of crewai, pydantic, langchain, or Python itself. I’ve seen this with environments pinned to older interpreter versions and half-upgraded packages.

Check your stack:

python --version
pip show crewai langchain pydantic anyio asyncio-run-in-threadpool 2>/dev/null
pip freeze | grep -E "crewai|langchain|pydantic|anyio"

A safe move is to upgrade to a supported Python version and reinstall dependencies cleanly:

pip install -U crewai langchain pydantic anyio

How to Debug It

1. •

   Read the exact traceback

   • •If you see RuntimeError: This event loop is already running, you’re nesting loops.
   • •If you see coroutine was never awaited, you forgot an await.
   • •If you see errors around Task, Crew, or tool execution, check where async boundaries are being crossed.
2. •

   Find every asyncio.run()

   • •Search your codebase for it.
   • •Remove it from notebooks, FastAPI routes, Streamlit callbacks, and any function that may already run inside an event loop.
3. •

   Trace where CrewAI is called

   • •Look at whether you’re calling:
     • •crew.kickoff()
     • •a custom async tool
     • •an LLM client directly inside a sync function
   • •The bug is usually one layer above where the stack trace points.
4. •

   Isolate with a minimal script

   • •Run CrewAI in a plain .py file with one agent and one task.
   • •If it works there but fails in your app, the issue is your host runtime, not CrewAI itself.

Example minimal test:

import asyncio
from crewai import Agent, Task, Crew

agent = Agent(role="Tester", goal="Verify execution", backstory="Tests things.")
task = Task(description="Say hello.", agent=agent)
crew = Crew(agents=[agent], tasks=[task])

async def main():
    print(await crew.kickoff())

asyncio.run(main())

If this works in terminal but fails in notebook or API server, stop debugging CrewAI internals and fix your runtime integration.

Prevention

• •Keep one rule: sync entrypoints use asyncio.run(), async entrypoints use await.
• •Don’t put event-loop management inside tools, agents, or task logic.
• •Test CrewAI flows in both:
  • •plain Python scripts
  • •your actual host runtime like FastAPI or Jupyter
• •Pin compatible versions of Python and dependencies before shipping to production.

Keep learning

• •The complete AI Agents Roadmap — my full 8-step breakdown
• •Free: The AI Agent Starter Kit — PDF checklist + starter code
• •Work with me — I build AI for banks and insurance companies

By Cyprian Aarons, AI Consultant at Topiax.