Ai Research Assistant

An implementation of a Model Context Protocol (MCP) server built in Python using the FastMCP framework. This server acts as an AI Research Assistant, providing tools, resources, and prompt templates to help LLM clients (like Claude Desktop) read, write, and summarize research notes stored in a local flat-file database.

---

What the Project Does

This MCP server exposes the following capabilities to any compatible LLM client:

1. Tools (tools)

• add_research(research: str): Appends a new line of research notes or text to the local database file (research.txt).
• read_research(): Reads and returns the entire contents of the research database. If the file is empty, it returns a message indicating no research has been saved yet.

2. Resources (resources)

• research://latest: A dynamic URI resource that retrieves only the latest research entry (the last line of the research.txt file).

3. Prompts (prompts)

• research_summary_prompt: A prompt template that reads the current contents of research.txt and automatically formats a prompt asking the AI model to summarize the collected research.

---

Directory Structure

• main.py: The entry point of the MCP server implementing the tools, resources, and prompts using FastMCP.
• research.txt: The local storage file containing the research notes.
• pyproject.toml: The Python project configuration defining metadata and dependencies (mcp[cli]).
• uv.lock: The lockfile for deterministic dependency resolution via uv.

---

Setup Instructions

Prerequisites

• Python: Version 3.12 or higher (configured via .python-version).
• uv: It is highly recommended to use uv for fast dependency management and running the server. If you don't have it, install it using:

  curl -LsSf https://astral.sh/uv/install.sh | sh
  

Installation

1. Clone or navigate to the project directory:

   cd /Users/gaganchaudhary/mcp-server-demo
   

2. Create the virtual environment and install dependencies:

   uv sync
   

---

Running the Server

1. Developer Inspection & Testing (Recommended)

To test and interact with the server interactively using the MCP Inspector web interface, run:

uvx mcp dev main.py

This command starts the server and hosts a visual inspector tool locally (typically at http://localhost:5173) where you can trigger tools, read resources, and test prompts.

2. Standard Run Command

To run the server directly on standard input/output (stdio) transport:

uv run main.py

---

Client Integration

To integrate this MCP server with Claude Desktop, add it to your configuration file.

Configuration File Location

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Configuration Content

Open the configuration file and add the ai-research-assistant server under the mcpServers object:

{
  "mcpServers": {
    "ai-research-assistant": {
      "command": "uv",
      "args": [
        "--directory",
        "/Users/gaganchaudhary/mcp-server-demo",
        "run",
        "main.py"
      ]
    }
  }
}

After modifying the configuration file, restart Claude Desktop. You will see the new hammer icon indicating that the AI Research Assistant tools are available!