Composio Agent Builder

Table of Contents

• Overview
• Core Features
• Architecture Overview
• Authentication
• Dashboard Modules
• Tech Stack
• Example Use Case
• Installation & Setup
  • Clone the repository
  • Environment Variables
  • Docker Setup
  • Manual Setup
    • Backend Setup
    • Frontend Setup (dashboard)

Overview

This project enables autonomous task execution through an AI-powered agent that can analyze a user’s task description, derive the required tools and toolkits, design an execution plan, and implement it automatically using Composio’s Tool Router.

It provides two major modes of interaction:

1. Chat Interface – Interactively communicate with the agent to accomplish tasks.
2. Scheduled Agent Runs (Cron) – Automate workflows to run at specified times.

All settings, authentications, and configurations are managed via a user-friendly dashboard.

Core Features

1. Task Understanding & Planning

• The user inputs a task description.
• The system analyzes the task to:
  • Identify relevant tools and toolkits.
  • Design a step-by-step plan for implementation.

2. Dynamic Tool Integration via Composio

• The project leverages Composio’s Tool Router to:
  • Integrate and manage all required tools dynamically.
  • Route execution through appropriate services.
• The agent automatically connects and executes using authenticated tools.

3. Chat Interface

• Chat directly with your intelligent agent.
• Discuss, modify, and monitor tasks in real time.
• Great for debugging or performing one-off task executions.

4. Scheduled Agent Runs (Cron Jobs)

• Set time-based workflows to automate task execution.
• Ideal for repetitive or recurring operations.

5. Assistant Dashboard

• Unified control panel for:
  • Chat Interface
  • Scheduled Agent Runs
  • Tool Authentication & Access Management
• Provides seamless configuration and monitoring.

6. Exporting and Importing Agents

• Easily import agents from json format.
• Share your custom-built agents with others by exporting them.

Architecture Overview

flowchart TD
    A[User Input: <br>Task Description ] --> B[Task Analyzer]
    B --> C[Tool & Plan Derivation]
    C --> D[Composio Tool Router]
    D --> E[Agent Design & <br>Integration]
    E --> F{Execution via}
    F --> G[Chat Interface: Real-time Interaction]
    F --> H[Scheduled Runs: <br>Automated Cron Jobs]
    G --> I[Dashboard: <br>Authentication, <br>Monitoring]
    H --> I

Loading

Authentication

• Each integrated tool requires user authentication.
• Authentication is handled securely through the assistant dashboard.
• Once authorized, the agent gains controlled access to perform operations using that tool.

Dashboard Modules

Module	Description
Chat Interface	Converse with the agent to execute or test tasks in real-time.
Scheduled Runs	Configure cron-like schedules for periodic task execution.
Tool Authentication	Securely connect third-party tools.

Tech Stack

• LLM / AI Frameworks: OpenAI / Langgraph framework
• Tool Integration: Composio Python SDK
• Backend: Python / FastAPI
• Frontend: React / Next.js
• Scheduling: Cron
• Database: SQLite

Example Use Case

Task Description:

“Fetch latest news of the given company and draft the summary in google docs”

Created Plan (using Tool Router):

1. Use the COMPOSIO_SEARCH_NEWS tool to search for the most recent news articles related to the given company. Use a relevant time filter such as the past day or week to ensure up-to-date news.
2. Analyze and summarize the retrieved news articles to create a clear, informative summary.
3. Use the GOOGLEDOCS_CREATE_DOCUMENT_MARKDOWN tool to create a new Google Docs document titled as specified by the user, and populate it with the summary content in markdown format.

Instructions: Strictly follow the above plan. Use only the specified tools for searching news and creating the Google Docs document. Request user clarification only if completely blocked due to missing or unclear information.

Tools: ["COMPOSIO_SEARCH_NEWS", "GOOGLEDOCS_CREATE_DOCUMENT_MARKDOWN"]

Installation & Setup

Clone the repository

git clone https://github.com/OSSML/Composio_Agent_Builder.git
cd Composio_Agent_Builder

Environment Variables

Copy .env.example to .env and configure values:

• Get your OpenAI api key from here
• Get your Composio api key from here

cp .env.example .env

COMPOSIO_API_KEY= "YOUR_API_KEY"
OPENAI_API_KEY= "YOUR_API_KEY"

# ----- Optional ENV variables -----

## Optional Database file path (Must be empty for docker build)
## Saves the .db file in the project root directory if not specified.
## Format f"sqlite+aiosqlite:///{.db posix file path}"
# DATABASE_URL=...

### Logging configuration
## default "LOCAL". Anything other than "LOCAL" or "DEVELOPMENT" will output structured logging (suitable for filtering) instead of string logging
# ENV_MODE=...
## default "INFO"
# LOG_LEVEL=...

## Optional Langsmith tracing
# LANGSMITH_TRACING=true
# LANGSMITH_ENDPOINT=https://api.smith.langchain.com
# LANGSMITH_API_KEY=...
# LANGSMITH_PROJECT=composio-agent-builder

Docker Setup

docker-compose up

Manual Setup

Backend Setup

cd backend

Using uv (recommended)

# Install dependencies
uv sync
# Run the Backend
uv run main.py

Using pip

# creating a virtual environment
python -m venv env
# Activation for linux
source env/Scripts/activate
# Activation for windows
env/Scripts/activate.ps1

# Install dependencies
pip install -r pyproject.toml

# Run the Backend
python main.py

Frontend Setup (dashboard)

# In new terminal
cd frontend
npm install
npm run dev

Then open the dashboard at: http://localhost:3000

The project comes with 3 default agents to experiment with.

License

Apache 2.0 License - see LICENSE file for details.