Docbot

Bidirectional CSV ↔️ JSON ↔️ Markdown transformer exposed as an MCP (Model Context Protocol) server with HTTP transport.

Overview

docbot-mcp is a Model Context Protocol server that provides tools for converting between CSV, JSON, and Markdown formats. It's built on the Model Context Protocol and designed for deployment on serverless platforms like AWS App Runner.

Features

• ✅ CSV to JSON — Parse CSV strings into JSON arrays
• ✅ JSON to CSV — Serialize JSON arrays to CSV format
• ✅ CSV to Markdown — Convert CSV to formatted Markdown tables
• ✅ JSON to Markdown — Convert JSON arrays to Markdown tables
• ✅ HTTP-based — REST API endpoint for easy integration
• ✅ Containerized — Ready for deployment on AWS App Runner, Docker, etc.

Installation

npm install

Requirements

• Node.js 16+ (ES modules support)
• Dependencies:
  • @modelcontextprotocol/sdk — MCP protocol implementation
  • @rflukerii/docbot — Document conversion utilities
  • express — HTTP server framework
  • zod — Input validation

Usage

Starting the Server

npm start

The server starts on http://localhost:3000 with the MCP endpoint at /docbot-mcp.

Available Tools

All tools accept input as text parameters and return formatted text responses.

csvToJson

Converts a CSV string to JSON format.

Input: csv (string)
Output: JSON array as string

curl -X POST http://localhost:3000/docbot-mcp \
  -H "Content-Type: application/json" \
  -d '{"tool": "csvToJson", "input": {"csv": "name,age\nAlice,30"}}'

jsonToCsv

Converts a JSON array to CSV format.

Input: json (string)
Output: CSV string

curl -X POST http://localhost:3000/docbot-mcp \
  -H "Content-Type: application/json" \
  -d '{"tool": "jsonToCsv", "input": {"json": "[{\"name\":\"Alice\",\"age\":30}]"}}'

csvToMarkdown

Converts a CSV string to a Markdown table.

Input: csv (string)
Output: Markdown table string

jsonToMarkdown

Converts a JSON array to a Markdown table.

Input: json (string)
Output: Markdown table string

Architecture

HTTP Transport

The server uses StreamableHTTPServerTransport instead of stdio for better compatibility with containerized environments. Each request to /docbot-mcp creates a new server instance, processes the MCP request, and returns the result.

Express Server (port 3000)
    ↓
POST /docbot-mcp endpoint
    ↓
MCP Server instance (createServer)
    ↓
StreamableHTTPServerTransport
    ↓
Tool execution (csvToJson, jsonToCsv, etc.)
    ↓
Response back to client

File Structure

• index.js — Main HTTP server entry point (AWS App Runner compatible)
• index-stdio.js — Stdio transport version (for local/stdio-based clients)
• package.json — Dependencies and scripts

Deployment

AWS App Runner

The server includes a start script configured for AWS App Runner:

{
  "scripts": {
    "start": "node index.js"
  }
}

Steps to deploy:

1. Push code to AWS CodeCommit, GitHub, or BitBucket
2. Create AWS App Runner service
3. Configure source as your repository
4. Set build command: npm install
5. Set start command: npm start
6. App Runner will automatically run the server on port 3000

Docker

Example Dockerfile:

FROM node:18-alpine
WORKDIR /app
COPY package*.json ./
RUN npm install --production
COPY . .
EXPOSE 3000
CMD ["npm", "start"]

Development

Switching Between Transports

• HTTP (default): node index.js — For containerized environments
• Stdio: node index-stdio.js — For direct CLI/stdio clients

Adding New Tools

Edit index.js and use the server.registerTool() pattern:

server.registerTool(
  'toolName',
  {
    title: 'Display Title',
    description: 'What it does',
    inputSchema: { paramName: z.type() }
  },
  async ({ paramName }) => ({
    content: [{ type: 'text', text: 'result' }]
  })
);

License

MIT

Changelog

See CHANGELOG.md for version history and updates.