# Model Context Protocol (MCP)

## Modelcontextprotocol.io

- [docs](https://modelcontextprotocol.io/introduction)
- [Building effective agents](https://www.anthropic.com/engineering/building-effective-agents)
- [python-sdk source](https://github.com/modelcontextprotocol/python-sdk)

## Reference Implementations

### FastMCP

- [source](https://github.com/jlowin/fastmcp)
    - [fork](git@github.com:wgong/fastmcp.git) : (~/projects/wgong/MCP/fastmcp)
- [docs](https://gofastmcp.com/getting-started/welcome)

A high-level, fast, Pythonic way to build MCP servers and clients

```bash
# install
pip install fastmcp

# verify version
fastmcp version

FastMCP version:                                                            2.5.2
MCP version:                                                                1.9.2
Python version:                                                            3.13.2
Platform:                            Linux-6.8.0-60-generic-x86_64-with-glibc2.35
FastMCP root path:              ~/anaconda3/envs/mcp/lib/python3.13/site-packages

# dev install
git clone https://github.com/jlowin/fastmcp.git
cd fastmcp
uv sync

```

### Fast Agent

- [source](https://github.com/evalstate/fast-agent)
    - [fork](https://github.com/wgong/fast-agent) : (~/projects/wgong/MCP/fast-agent)
- [story](https://llmindset.co.uk/resources/fast-agent/)

A framework with complete, end-to-end tested MCP Feature support including Sampling, which helps create and interact with sophisticated Agents and Workflows in minutes. 
Both Anthropic (Haiku, Sonnet, Opus) and OpenAI models (gpt-4o/gpt-4.1 family, o1/o3 family) are supported


## Demos

### FastMCP 

A `fastmcp` tutorial

- MCP server:
    - `mcp_server.py`: collection of tools/resources like calculator, trigonometry function, yahoo finance API call
- MCP clients:
    - `mcp_client_simple.py`: simple client with rule-based query parsing
    - `mcp_client_llm.py`: client with LLM-based query parsing
    - `mcp_client_llm_resource.py`: client with LLM parsing and calling tool/resource (see `readme_tool_resource.md` for details)

see [readme_gemini.md](https://github.com/digital-duck/mcp_demo/blob/main/readme_gemini.md) for more details

#### Setup
```bash
conda create -n mcp
conda activate mcp
git clone https://github.com/digital-duck/mcp_demo.git

cd mcp_demo/fastmcp
pip install -r requirements.txt

# in 1st terminal
python mcp_server.py

# in 2nd terminal
python mcp_client_llm_resource.py
# python mcp_client_simple.py
# python mcp_client_llm.py
```

#### Streamlit MCP app

see `fastmcp/README.md`

```bash
streamlit run st_mcp_rag.py
```

#### FastMCP - AWS 

**To Be Built**

see [readme_aws.md](https://github.com/digital-duck/mcp_demo/blob/main/readme_aws.md) for more details


### MCP SDK 

see `mcp_sdk` sub-folder

**To Be Verified**