A collection of 50+ runnable Jupyter notebook tutorials covering GenAI agent techniques, from simple chatbots to multi-agent systems.
GoogleCloudPlatform/cymbal-air-toolbox-demo
A demo customer service agent for a fictional airline, built with LangGraph and RAG to query Google Cloud databases through MCP Toolbox.
- Stars
- 350
- Forks
- 123
- Open issues
- 53
- Last commit
- 3 months ago
What is it?
This project is a reference implementation of an agentic customer service application built on Google Cloud databases. It uses a fictional airline, Cymbal Air, and its hub at San Francisco International Airport as the scenario. The assistant can help travelers manage flights and answer questions about the airport, such as nearby coffee, terminal shops, and flights to a given city.
The agent is orchestrated with LangGraph, so the LLM chooses which tools to call and in what order rather than following a fixed chain of calls. Each tool maps to a predefined database query, such as finding flights or listing airport amenities. Retrieval Augmented Generation grounds the answers in data stored in the database.
Database access goes through MCP Toolbox, a middleware server that exposes database operations as tools and handles authentication and authorization, so the agent never connects to the database directly. The database layer is described as interchangeable, and the demo includes setup for Cloud SQL-compatible and other Google Cloud databases through a tools.yaml configuration.
Key features
- 01Agent orchestration with LangGraph where the LLM selects tools dynamically.
- 02Database access through MCP Toolbox, which handles authentication and authorization.
- 03Retrieval Augmented Generation to ground answers in database data.
- 04Predefined tools such as find_flights and list_amenities mapped to database queries.
- 05Configurable tools via a tools.yaml file, runnable locally or on Cloud Run.
Use cases
- Learning how to build a customer service agent that queries a cloud database.
- Starting point for a custom agentic app that uses MCP Toolbox with your own database.
- Exploring how to restrict agent database access through a middleware layer.
Quick start
- 1
Clone the repository
git clone https://github.com/GoogleCloudPlatform/cymbal-air-toolbox-demo.git - 2
Download the MCP Toolbox binary (Linux amd64 example from the README)
curl -O https://storage.googleapis.com/genai-toolbox/v$VERSION/linux/amd64/toolbox - 3
Complete the one-time database and tools.yaml setup using the Database Setup Guide
- 4
Run the Toolbox server locally
./toolbox --tools-file "tools.yaml"
Things to know
- The README states this is a demonstration and not an officially supported Google product.
- Running it requires setting up a Google Cloud database and the MCP Toolbox server first.
- The setup is multi-step and depends on the Database Setup Guide and separate app run instructions.
In collections
Topics
FAQ
- What is cymbal-air-toolbox-demo?
- A demo customer service agent for a fictional airline, built with LangGraph and RAG to query Google Cloud databases through MCP Toolbox. This project is a reference implementation of an agentic customer service application built on Google Cloud databases. It uses a fictional airline, Cymbal Air, and its hub at San Francisco International Airport as the scenario. The assistant can help travelers manage flights and answer questions about the airport, such as nearby coffee, terminal shops, and flights to a given city.
- Is cymbal-air-toolbox-demo free to use? What license does it have?
- cymbal-air-toolbox-demo is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
- Is cymbal-air-toolbox-demo still maintained?
- Updated within 6 months. The last commit to cymbal-air-toolbox-demo was on Aug 14, 2026.
- What can I use cymbal-air-toolbox-demo for?
- Learning how to build a customer service agent that queries a cloud database. Starting point for a custom agentic app that uses MCP Toolbox with your own database. Exploring how to restrict agent database access through a middleware layer.
- How do I install cymbal-air-toolbox-demo?
- Clone the repository: git clone https://github.com/GoogleCloudPlatform/cymbal-air-toolbox-demo.git Download the MCP Toolbox binary (Linux amd64 example from the README): curl -O https://storage.googleapis.com/genai-toolbox/v$VERSION/linux/amd64/toolbox Complete the one-time database and tools.yaml setup using the Database Setup Guide Run the Toolbox server locally: ./toolbox --tools-file "tools.yaml"
- Where can I get cymbal-air-toolbox-demo?
- The source code and installation instructions are on GitHub: https://github.com/GoogleCloudPlatform/cymbal-air-toolbox-demo
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
An opinionated, batteries-included agent harness for Python that runs long multi-step tasks out of the box on LangGraph.
A collection of 28 code-first tutorials for taking GenAI agents from prototype to production, covering memory, RAG, deployment, security and observability.
Open-source, self-hostable AI-native backend that gives AI-generated apps a database, auth, storage, memory, functions and CDN hosting in one service.
Agent memory system that stores, recalls and reflects on memories so AI agents learn over time instead of only replaying chat history.
A tutorial repo of five Jupyter notebooks that walk through building a LangGraph-based deep research agent from scratch.