使用说明
Google ADK, Gemini, and MCP Tool Integration
Interactive Streamlit app
Explore streamlit_app/app.py, an owner-configured MCP tool inspector. Live tools require an owner-approved HTTPS Streamable HTTP server plus an explicit tool allowlist. No live server was configured or tested.
Run with python -m streamlit run streamlit_app/app.py after installing streamlit_app/requirements.txt. See DEPLOYMENT.md for deployment and access requirements. A live deployment is not yet confirmed.
A notebook demonstrating agent setup, conversation sessions, retry settings, and an MCP toolset connection.
Why this project matters
Tool integration lets an agent operate beyond text generation; session management and error handling make that integration easier to reason about.
Technical focus
Python · Google ADK · Gemini · MCP · stdio tool transport · session identifiers
Explore the implementation
Open the notebook to inspect preprocessing, implementation, and saved outputs.
Run locally
git clone https://github.com/Arjun1443/ADK-Google-ADK-Gemini-MCP-Integration.git
cd ADK-Google-ADK-Gemini-MCP-Integration
python -m venv .venv
# Activate .venv before continuing.
python -m pip install google-adk google-genai jupyter
python -m jupyter notebook
Select ADK-3 Google ADK + Gemini + MCP Integration.ipynb in Jupyter and review its configuration before executing cells in order.
Reproducibility notes
The notebook uses Kaggle Secrets; adapt credential loading if running outside Kaggle. Its stdio MCP example also requires Node.js and npx. Review the external server command before execution. This is an integration demonstration, not a validated production deployment. No LICENSE file is present in the inspected tree; no new license is granted by this README.
Saved notebook outputs describe a prior execution; they are not an independently reproduced benchmark. Dependency versions are not locked, so compatibility may require adjustment.
Author
Naga Anweesh Reddy · AI development and technical training
