KONTEKST MCP Server

MCP server for KONTEKST - access your context files from Claude Desktop and other MCP-compatible AI tools.

Installation

npm install -g @kontekst/mcp-server

Configuration

Add to your Claude Desktop config (%APPDATA%\Claude\claude_desktop_config.json):

{
  "mcpServers": {
    "kontekst": {
      "command": "npx",
      "args": ["@kontekst/mcp-server"],
      "env": {
        "KONTEKST_API_TOKEN": "your-api-token-here",
        "KONTEKST_API_URL": "https://kontekst.vercel.app"
      }
    }
  }
}

Getting Your API Token

1. Visit kontekst.vercel.app
2. Go to Settings
3. Generate an API token

Available Tools

| Tool | Description | |------|-------------| | list_files | List all your KONTEKST context files | | read_file | Read a specific file by path | | write_file | Write or update a file |

Usage

Once configured, you can use natural language to interact with your context files:

• "List my context files"
• "Read my me.md file"
• "Write a new note about project status to notes/status.md"

Troubleshooting

If you get permission errors, make sure:

1. Your API token is valid and not expired
2. The token has the necessary permissions for the operations you're trying to perform

HTTP Mode (Remote MCP)

For connecting AI tools via HTTP instead of stdio:

# Run as HTTP server locally
MCP_MODE=http KONTEKST_API_TOKEN=your-token npx @kontekst/mcp-server

This starts an Express server at http://localhost:3001/mcp.

Deploying to Vercel (Server-side MCP)

For hosting MCP on Vercel for others to use, add a new API route pages/api/mcp.ts to your Next.js app:

// pages/api/mcp.ts
// Implements MCP JSON-RPC protocol over HTTP

This creates https://your-app.vercel.app/api/mcp that users can connect their AI tools to.