# MCP Chat CLI

A command-line chat interface that connects to an MCP server using the Anthropic API. Built while working through Anthropic's *Introduction to Model Context Protocol* course, extended with custom tools, resources, and prompts.

## What this is

MCP (Model Context Protocol) is an open standard for connecting AI models to external tools and data sources. This project implements both sides of that connection — a FastMCP server that exposes documents as resources and defines tools for reading and editing them, and a client that connects to the server and makes those capabilities available inside a chat interface.

The server defines:

- **Tools** — read and edit documents
- **Resources** — list all documents or fetch a specific one by URI
- **Prompts** — reformat a document to markdown, or summarize its contents

The client implements the full MCP client session, including tool calls, resource reads, prompt retrieval, and command autocompletion.

## What I worked through

Starting from a course starter pack, I implemented the missing pieces on both sides:

- `read_resource`, `list_prompts`, and `get_prompt` on the client
- Resource endpoints (`docs://documents` and `docs://documents/{doc_id}`) on the server
- Two prompts (`format` and `summarize`) that instruct the model to use the available tools

The main thing this project made concrete for me is the separation between the server (which defines what's available) and the client (which knows how to call it) — and how prompts are just structured messages that give the model a starting context, not magic.

## Prerequisites

- Python 3.9+
- Anthropic API key

## Setup

1. Clone the repo and navigate into the project folder.

2. Create a virtual environment and activate it:

```bash
uv venv
.venv\Scripts\activate  # Windows
source .venv/bin/activate  # Mac/Linux
```

1. Install dependencies:

```bash
uv pip install -e .
```

1. Create a `.env` file in the project root:

```
ANTHROPIC_API_KEY="your-key-here"
```

1. Run the app:

```bash
uv run main.py
```

## Usage

Type a message to chat. Use `@doc_id` to include a document in your query, and `/command` to trigger a prompt. Tab autocompletes available commands.

```
> Tell me about @deposition.md
> /summarize report.pdf
> /format plan.md
```

To add your own documents, edit the `docs` dictionary in `mcp_server.py`.

## Testing the server directly

```bash
mcp dev mcp_server.py
```

This opens the MCP Inspector in your browser where you can test tools, resources, and prompts without the chat interface.

## Project structure

```
mcp_chat_cli/
├── main.py              # entrypoint
├── mcp_server.py        # FastMCP server — tools, resources, prompts
├── mcp_client.py        # MCP client session wrapper
├── core/
│   ├── chat.py          # chat loop logic
│   ├── claude.py        # Anthropic API integration
│   ├── cli.py           # CLI setup and input handling
│   ├── cli_chat.py      # connects CLI and chat
│   └── tools.py         # tool call handling
├── .env                 # API key (not committed)
└── pyproject.toml       # dependencies
```