# Akib Dabgar MCP Server

A production-ready, TypeScript-based **Model Context Protocol (MCP)** server built for **Akib Dabgar**. The server acts as a structured personal knowledge hub, allowing LLM clients (such as Claude Desktop, ChatGPT, Claude Code, and Codex) to query details about Akib's skills, experience, projects, education, and career aspirations.

To maximize context-window efficiency and execution speed, the server implements **Dynamic Context Retrieval (DCR)**, selectively parsing and returning only relevant sections of Akib's profile knowledge base instead of dumping the entire profile into the context.

---

## 🚀 Features

- **Dynamic Context Retrieval (DCR):** Implements keyword tokenization and weight-based section scoring to rank and retrieve only the most relevant headings/sections matching the query.
- **MCP Tools:** Exposes a `query_profile` tool that takes natural language queries and returns optimized Markdown snippets.
- **MCP Resources:** Maps all markdown files in the `knowledge/` database to `knowledge://` URIs, enabling client LLMs to read raw content when needed.
- **MCP Prompts:** Exposes an `akib_profile_assistant` prompt that pre-loads system prompt instructions, response policies, and optional topic-specific retriever context.
- **TypeScript ESM Architecture:** Clean, modular, typed codebase utilizing modern Node.js and ES Modules.

---

## 📁 Folder Structure

```text
├── knowledge/
│   ├── dcr/                 # DCR architecture specifications
│   │   ├── overview.md
│   │   ├── retrieval-strategy.md
│   │   ├── ranking-rules.md
│   │   ├── query-examples.md
│   │   ├── response-policy.md
│   │   └── limitations.md
│   ├── about.md             # Core bio, location, professional summary
│   ├── skills.md            # Technical and soft skills
│   ├── projects.md          # Akib's project history (ChainIT Pay, etc.)
│   ├── experience.md        # Work experience details (eSparkBiz)
│   ├── education.md         # Saurashtra University degree
│   ├── career-goals.md      # Professional goals
│   ├── interests.md         # Architecture, scalable systems, AI
│   └── faq.md               # Common questions & answers
├── src/
│   ├── retriever/
│   │   └── DynamicContextRetriever.ts # DCR query analysis and section ranking
│   ├── services/
│   │   └── KnowledgeService.ts        # Markdown loader and segment parser
│   ├── tools/
│   │   └── index.ts                   # Tool schema definitions and handlers
│   └── index.ts                       # Server entry point, transport & handlers
├── package.json
├── tsconfig.json
└── README.md                # Project documentation (this file)
```

---

## 🛠️ Getting Started

### Prerequisites

- [Node.js](https://nodejs.org/) (v18.0.0 or higher recommended)
- npm (v9.0.0 or higher)

### Installation

Clone the repository and install the dependencies:

```bash
npm install
```

### Build

Compile the TypeScript codebase to the `build/` directory:

```bash
npm run build
```

---

## ⚙️ Running the Server

Start the MCP Server on standard I/O (Stdio):

```bash
npm start
```

For development with hot-reloading:

```bash
npm run dev
```

---

## 🔌 Client Configuration

To connect this server to commonly used MCP clients:

### 1. Claude Desktop
Add the server configuration to your `claude_desktop_config.json` (located at `%APPDATA%\Claude\claude_desktop_config.json` on Windows or `~/Library/Application Support/Claude/claude_desktop_config.json` on macOS):

```json
{
  "mcpServers": {
    "akib-profile-server": {
      "command": "node",
      "args": [
        "d:/text/jpg/Projects/knowledge/build/index.js"
      ],
      "env": {}
    }
  }
}
```

### 2. Cursor IDE
1. Open Cursor Settings -> **Features** -> **MCP**.
2. Click **+ Add New MCP Server**.
3. Fill in the configuration:
   - **Name:** `Akib Profile Server`
   - **Type:** `stdio`
   - **Command:** `node d:/text/jpg/Projects/knowledge/build/index.js`
4. Click **Save**.

### 3. Claude Code (CLI)
Add the server using the `mcp` command:
```bash
claude mcp add akib-profile-server node d:/text/jpg/Projects/knowledge/build/index.js
```

---

## 🔍 How to Test

### Using the MCP Inspector

You can test the server interactively by running the MCP Inspector:

```bash
npx @modelcontextprotocol/inspector node build/index.js
```

### Manual Verification via Stdio

You can verify the JSON-RPC interface by piping requests to stdin:

1. **List Tools:**
   ```bash
   echo {"jsonrpc":"2.0","method":"tools/list","id":1,"params":{}} | node build/index.js
   ```

2. **Query Profile (DCR in action):**
   ```bash
   echo {"jsonrpc":"2.0","method":"tools/call","id":2,"params":{"name":"query_profile","arguments":{"query":"What databases does Akib know?"}}} | node build/index.js
   ```

3. **List Prompts:**
   ```bash
   echo {"jsonrpc":"2.0","method":"prompts/list","id":3,"params":{}} | node build/index.js
   ```

4. **Get Prompt:**
   ```bash
   echo {"jsonrpc":"2.0","method":"prompts/get","id":4,"params":{"name":"akib_profile_assistant","arguments":{"topic":"skills"}}} | node build/index.js
   ```

---

## 💡 Architectural Highlights

- **Granular Parsing:** `KnowledgeService` splits markdown files by headings, parsing them into indexable `KnowledgeSection` blocks. This ensures high-granularity DCR mapping.
- **Smart Token & Synonym Routing:** `DynamicContextRetriever` tokenizes query keywords, strips stopwords, and uses a keyword-to-file boost matrix combined with bidirectional heading/content string matches.
- **Robust Path Security:** Target resource URIs are normalized and restricted within the root `knowledge/` directory to prevent directory traversal vulnerabilities.