How to intermediate · 3 min read

How to create custom tool for CrewAI

Q: How to create custom tool for CrewAI

To create a custom tool for CrewAI, build a Python API or function that exposes your tool's capabilities and register it within the CrewAI platform. Use CrewAI's SDK or REST API to integrate your tool, enabling CrewAI to call it dynamically during workflows.

Quick answer

To create a custom tool for CrewAI, build a Python API or function that exposes your tool's capabilities and register it within the CrewAI platform. Use CrewAI's SDK or REST API to integrate your tool, enabling CrewAI to call it dynamically during workflows.

PREREQUISITES

Python 3.8+
CrewAI account with API access
pip install requests
API key or credentials for CrewAI

Setup

Install the requests library to interact with CrewAI's API and set your environment variables for authentication.

Run pip install requests to install dependencies.
Export your CrewAI API key: export CREWAI_API_KEY='your_api_key' on Unix or set environment variable on Windows.

bash

pip install requests

Step by step

This example shows how to create a simple custom tool that CrewAI can call via its API. The tool is a Python Flask app exposing an endpoint CrewAI invokes.

python

import os
from flask import Flask, request, jsonify
import requests

app = Flask(__name__)

CREWAI_API_KEY = os.environ.get('CREWAI_API_KEY')
CREWAI_API_URL = 'https://api.crewai.com/v1/tools/register'

# Define your custom tool logic
@app.route('/custom-tool', methods=['POST'])
def custom_tool():
    data = request.json
    # Example: echo input text reversed
    input_text = data.get('text', '')
    result = input_text[::-1]
    return jsonify({'result': result})

# Register the tool with CrewAI
def register_tool():
    headers = {'Authorization': f'Bearer {CREWAI_API_KEY}', 'Content-Type': 'application/json'}
    payload = {
        'name': 'ReverseTextTool',
        'description': 'Reverses input text',
        'endpoint': 'https://your-server.com/custom-tool',
        'method': 'POST',
        'input_schema': {'type': 'object', 'properties': {'text': {'type': 'string'}}},
        'output_schema': {'type': 'object', 'properties': {'result': {'type': 'string'}}}
    }
    response = requests.post(CREWAI_API_URL, json=payload, headers=headers)
    if response.status_code == 200:
        print('Tool registered successfully')
    else:
        print('Failed to register tool:', response.text)

if __name__ == '__main__':
    register_tool()
    app.run(host='0.0.0.0', port=5000)

output

Tool registered successfully
 * Running on http://0.0.0.0:5000/ (Press CTRL+C to quit)

Common variations

You can implement your custom tool asynchronously using FastAPI or add authentication to your endpoint. Also, CrewAI supports different HTTP methods and payload formats.

python

from fastapi import FastAPI, Request
import uvicorn

app = FastAPI()

@app.post('/custom-tool')
async def custom_tool(request: Request):
    data = await request.json()
    input_text = data.get('text', '')
    return {'result': input_text.upper()}

if __name__ == '__main__':
    uvicorn.run(app, host='0.0.0.0', port=8000)

output

INFO:     Started server process [12345]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)

Troubleshooting

If tool registration fails, verify your API key and endpoint URL.
Ensure your server is publicly accessible for CrewAI to call your tool.
Check logs for HTTP errors and fix JSON schema mismatches.

✅

Key Takeaways

Use a web-accessible API endpoint to expose your custom tool for CrewAI integration.
Register your tool with CrewAI via their API including endpoint and schema details.
Test your tool locally and ensure it handles expected input/output JSON formats.
Secure your tool endpoint and validate API keys to prevent unauthorized access.

Verified 2026-04

Verify ↗