This project demonstrates an outreach automation workflow using Google ADK and Composio Gemini. It features two main AI agents:

• BDR Agent: Performs research on a candidate and company, then drafts a hyper-personalized outreach email using the research and Gmail tools.
• Sales Agent: Gathers detailed information about a contact and company using web, LinkedIn, and Google search tools, returning a comprehensive research report with sources.

The agents are orchestrated in a test script (main.py) that sends a sample query and prints the results.

.
├── main.py                  # Entrypoint script to run agent workflows
├── requirements.txt         # Python dependencies
├── custom_agents/
│   ├── ai_bdr_agent.py      # BDR Agent definition
│   └── ai_sales_agent.py    # Sales Agent definition
├── custom_tools/
│   └── tools.py             # Custom tool definitions for agents

• main.py: Sets up the session, initializes agents, and runs a test query. Handles printing and pretty-printing of agent responses and session state.
• custom_agents/ai_bdr_agent.py: Defines the BDR Agent, which uses the Sales Agent and Gmail tool to perform research and draft emails.
• custom_agents/ai_sales_agent.py: Defines the Sales Agent, which uses a variety of tools to gather and report on contact and company data.
• custom_tools/tools.py: Configures toolsets for the agents using Composio Gemini, including Gmail, EXA, and WebTool integrations.

1. Clone the repository and navigate to the project directory.

2. Install dependencies (use Python 3.10):

pip3.10 install -r requirements.txt

3. Set up environment variables:
   • Create a .env file in the project root with the following content:

COMPOSIO_API_KEY=your_composio_api_key_here

The script also sets GOOGLE_GENAI_USE_VERTEXAI=FALSE internally, so you do not need to set this manually.

4. Run the script:

python3.10 main.py

• The script loads environment variables and sets up a session using Google ADK's in-memory session service.
• It initializes the BDR Agent (which can call the Sales Agent and use Gmail) and the Sales Agent (which uses web and LinkedIn tools).
• The main() function sends a sample query (with contact and company info) to the BDR Agent.
• The BDR Agent performs research (via the Sales Agent), drafts an outreach email, and the script prints the results and session state.

• To add or modify agent logic, edit the files in custom_agents/.
• To add or configure tools, edit custom_tools/tools.py.
• The code is modular and can be extended with new agents or tools as needed.
• The agents use Google ADK's async runner and session management for scalable, stateful interactions.

• Python 3.10+
• API key for Composio Gemini (see https://composio.dev/ for details)