Building "Useful" Agents with Composio and CrewAI

This is a starter repository for building agents with Composio and CrewAI. You can use this codebase to update and build your own dynamic agents.

Usecase

This is an example for integrating Slack messages, conducting research, generating detailed analyses, and automating report creation in Notion.

Overview

Project execution follows this workflow:

• Using Composio to get Slack messages. Every time a Slack message is sent, a trigger is raised which is received by Composio and sent to the server.
• On Receiving the trigger, Crewai Agents are initiated
• Using CrewAI to build two agents: one for researching the topic and the second for writing a detailed analysis on the topic.
• Using Composio to write the detailed report in Notion, sending a confirmation message to Slack.

1. CrewAI powers the orchestration of the agents and tools.
2. Composio provides CrewAI the right tools which allows connecting to Slack (Receiving the message, sending a message to channel), Notion (Creating a page)

Here's a visual representation of the workflow:

Below is the basic setup you can do to run the repository.

Setup

1. Environment Setup

   Add a .env file and insert your OPENAI_API_KEY and Slack Channel ID. For Slack, identify the channel you wish to monitor. Navigate to the channel about section, and scroll to the bottom to locate the Channel ID.

   cp .env.sample .env

   Note: No need to add TRIGGER_ID in the .env file right now; we'll get to that in the next steps.

2. Run setup script

   This script installs necessary dependencies and tools for running the project.

   • Installs Poetry and project dependencies.
   • Installs Ngrok for tunneling.
   • Activates the Poetry shell for managing project dependencies.

   You can run this script to automate the setup process for your development environment.

   NB: Required Python version (>=3.10, <=3.13)

   chmod +x setup.sh && ./setup.sh

   Setup ngrok for tunneling and set up webhook (Use any method).

   • Using ngrok token

     ngrok config add-authtoken < your auth token >

   • Local ngrok setup

     ngrok http 2000

3. Composio Integration

   Use Composio CLI to add integrations for Notion and Slack:

   poetry run composio add notion
   poetry run composio add slack

   Complete the authorization flows for Notion and Slack.

4. Update the trigger callback URL

   poetry run composio triggers callback set "<ngrok-url>"

5. Enable Slack receive trigger

   poetry run composio triggers enable slack_receive_message

   Update TRIGGER_ID in .env

6. Run the Project

   To start the AI agents and initiate task execution:

   poetry run python main.py

   This command initializes the Client Crew, assembling the agents and assigning them tasks according to the configuration. By default, it generates a report.md file with the output of research on LLMs in the root folder.

   Start sending messages to the Slack channel and see Composio in action, seamlessly updating Notion with detailed reports.

Public Server Setup

For public setups, directly set the global trigger callback to the public URL and enable the trigger. To test out locally, you can use the command ngrok http PORT

Customization

You can modify this repository to create additional agents for various tasks:

1. Modify main.py and src/client/crew.py for new agents and tasks.
2. Update src/client/config/agents.yaml and src/client/config/tasks.yaml for new agents and tasks.
3. Adjust any other environment variables as needed.

Support

For support, questions, or feedback regarding the Composio platform,

• Check out our Documentation
• Join our Discord community