home / mcp / binal mcp server

Binal MCP Server

Provides MCP-based interactions between a web resume platform and Claude Desktop using Upstash Vector for RAG search.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "binal182-binal-mcpserver": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://your-app.vercel.app/api/mcp"
      ],
      "env": {
        "OPENAI_API_KEY": "sk-...",
        "UPSTASH_VECTOR_REST_URL": "https://ap-northeast-1.upstash.io/v1/vector",
        "UPSTASH_VECTOR_REST_TOKEN": "YOUR_UPSTASH_TOKEN"
      }
    }
  }
}

You set up an MCP (Model Context Protocol) server that enables Claude Desktop to interact with a web-based resume platform through a shared RAG search layer. This server connects your web app to an AI assistant, allowing interactive questions about the resume data and high-quality, vector-based searches for fast, contextual results.

How to use

You use an MCP client to connect to the server via the standard MCP transport. The server exposes a ready-made tool for querying Binal’s professional data and returning formatted results that the Claude Desktop assistant can present in conversations. When connected, you can ask questions about experience, skills, and projects, or trigger searches that return relevant resume snippets and context for deeper understanding.

How to install

Prerequisites: You need Node.js 18 or newer, the pnpm package manager, and an Upstash Vector Database account.

1) Prepare environment variables for the Upstash vector database and the AI integration.

UPSTASH_VECTOR_REST_URL="your_upstash_vector_url"
UPSTASH_VECTOR_REST_TOKEN="your_upstash_vector_token"
OPENAI_API_KEY="your_openai_api_key"

2) Install dependencies using pnpm.

pnpm install

3) Start the development server for local testing.

pnpm dev

4) Open http://localhost:3000 to view the web interface and test the MCP setup.

Configuration and MCP connection details

To enable Claude Desktop to connect to the MCP server, use the following configuration snippet. This runs the MCP bridge locally via npx and points to your deployed API. Replace the URL with your actual deployment endpoint.

{
  "mcpServers": {
    "binal_mcp": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://your-app.vercel.app/api/mcp"
      ]
    }
  }
}

Deployment notes and testing

If you plan to deploy to a hosting platform, ensure you provide the same environment variables to the runtime and update the MCP URL in the client configuration to the deployed endpoint. After deployment, restart Claude Desktop to establish the new MCP connection.

Security tip: keep your Upstash vector credentials and API keys secure. Do not expose them in public repos or client-side code. Use server-side environment variables and secret management where possible.

Troubleshooting

Common issues include connection failures between Claude Desktop and the MCP server, or missing environment variables. Verify that UPSTASH_VECTOR_REST_URL, UPSTASH_VECTOR_REST_TOKEN, and OPENAI_API_KEY are correctly set in your deployment environment and that the MCP endpoint URL is reachable from Claude Desktop.

Available tools

search_binal_knowledge

MCP tool for performing a semantic search over Binal’s professional data using the shared RAG logic and Zod-validated parameters.