This directory contains examples demonstrating how to use EnrichMCP.

-Available examples:

• hello_world - minimal "Hello, World" API
• hello_world_http - HTTP version using streamable HTTP
• shop_api - in-memory shop with relationships
• shop_api_sqlite - SQLite-backed shop
• sqlalchemy_shop - SQLAlchemy ORM version
• shop_api_gateway - gateway in front of FastAPI
• mutable_crud - mutable fields and CRUD decorators
• basic_memory - simple note-taking API using FileMemoryStore
• caching - request caching with ContextCache
• openai_chat_agent - interactive chat client
• server_side_llm_travel_planner - LLM-backed travel suggestions

The simplest EnrichMCP application with a single resource that returns "Hello, World!".

cd hello_world
python app.py

A variant of the Hello World example that serves the API over HTTP using the streamable HTTP transport.

cd hello_world_http
python app.py

Invoke the example using mcp_use:

python client.py

A comprehensive e-commerce API with multiple entities (users, orders and products).

cd shop_api
python app.py

This example demonstrates:

• Creating multiple entity models with rich descriptions
• Defining relationships between entities
• Page-based pagination for listing orders
• Fraud detection patterns and risk scoring
• In-memory data with filtering capabilities

A database-backed version of the shop API using SQLite.

cd shop_api_sqlite
python app.py

This example demonstrates:

• Database integration with SQLite
• Cursor-based pagination for efficient data streaming
• Lifespan management for database connections
• Dynamic sample data generation
• Real database queries with relationships

Both shop examples provide the same functionality but showcase different pagination strategies.

A version of the shop example built with SQLAlchemy ORM models. All entities are declared using SQLAlchemy and registered through include_sqlalchemy_models, which automatically creates CRUD endpoints and relationship resolvers. The sqlalchemy_lifespan helper manages the async engine, seeds the SQLite database on first run, and removes the file on shutdown when using cleanup_db_file=True.

To run this example:

cd sqlalchemy_shop
pip install -r requirements.txt
python app.py

This example demonstrates:

• Automatic conversion of SQLAlchemy models to EnrichMCP entities
• Auto-generated CRUD resources and relationship resolvers
• Async database access via SQLAlchemy
• Database seeding and pagination using the generated endpoints

A gateway example that forwards all requests to a FastAPI backend.

cd shop_api_gateway
uvicorn server:app --port 8001 &
python app.py

Stop the background server when finished. The gateway listens on port 8000 and provides the same schema-driven interface as the other examples.

A minimal API showcasing mutable fields and the new CRUD decorators.

cd mutable_crud
python app.py

Stop the background server when finished. The gateway listens on port 8000 and provides the same schema-driven interface as the other examples.

An interactive command-line chat agent that connects to one of the examples via MCP and lets you talk to it with either OpenAI or a local Ollama model.

cd openai_chat_agent
# install dependencies for the agent using uv
uv pip install -r requirements.txt
# copy the sample environment and optionally set OPENAI_API_KEY
cp .env.example .env
# run the chat agent
uv run app.py

Run the above commands from the openai_chat_agent directory so that config.json resolves the relative path to the shop_api example.

If OPENAI_API_KEY is not set the agent defaults to a local Ollama model defined by OLLAMA_MODEL (defaults to llama3). An Ollama server must be running locally when using this mode or the agent will fail to connect.

1. Install Ollama and ensure the ollama command is in your PATH.

2. Download the desired model (the example uses llama3.2 by default):

   ollama pull llama3.2

3. Start the Ollama server in the background before launching the chat agent:

   ollama serve &

The chat agent will fail to start if the server is not running. The included configuration starts the shop_api example using the MCP stdio connector so everything runs locally. This example demonstrates how to use MCPAgent with built-in conversation memory for chatting with your MCP data.