Tunzaa MCP Server

Grounding for AI-Driven Payment Integrations

The Tunzaa MCP Server is a developer-centric tool built for the community. It provides high-fidelity grounding data, integrated documentation, and "Golden" code patterns that allow AI agents (vibe coders) to generate perfect, non-hallucinated integration code for the Tunzaa ecosystem.

🚀 Instant Start (Fastest Way)

You can run the server directly from GitHub without cloning or installing dependencies.

1. Claude Desktop

Add this to your claude_desktop_config.json:

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

{
  "mcpServers": {
    "tunzaa": {
      "command": "npx",
      "args": ["-y", "github:Tunzaa/tunzaa_mcp"]
    }
  }
}

2. Cursor

1. Go to Settings -> Features -> MCP.
2. Click + Add New MCP Server.
3. Name: Tunzaa | Type: command | Value: npx -y github:Tunzaa/tunzaa_mcp

3. Windsurf

Add this to your ~/.codeium/config.json:

{
  "mcpServers": {
    "tunzaa": {
      "command": "npx",
      "args": ["-y", "github:Tunzaa/tunzaa_mcp"]
    }
  }
}

🧭 The Vibe Coding Workflow

This server is designed to help you build Tunzaa integrations in minutes. Follow this flow with your AI assistant:

1. Grounding: Add this MCP server to your project.
2. Exploration: Ask the AI: "List the Tunzaa resources and read the authentication guide."
3. Simulation: Run the tool: create_demo_shop to see a live trace of a successful integration.
4. Generation: Ask the AI: "Based on the grounding trace and the node-express example, build a checkout page for my app."

🏪 The Grounding "Demo Shop"

The create_demo_shop tool is the cornerstone of this platform. It doesn't just return data; it provides a Live Grounding Trace.

How to use it:

1. Trigger the Simulation: Tell your AI agent: "Run the Tunzaa create_demo_shop tool to understand the payment flow."
2. Review the Trace: The agent will receive a chronological sequence of calls including Authentication, Payment Initiation, and Installment creation.
3. Production Implementation: Each step in the trace contains "Grounding Insights" that teach the agent how to handle state, headers, and reference IDs in your actual code.
4. Boilerplate: Ask the agent to "Convert the grounding trace into a [Node/Python/PHP] implementation using the best practices found in the documentation resources."

✨ Features

• Integrated Documentation: AI agents can "read" guides on Auth, Payments, and Webhooks directly through MCP Resources.
• Golden Patterns: Embedded code snippets for Express.js, React Hooks, and more.
• Vibe Coder Optimized: Rich schema descriptions and instructional traces (via create_demo_shop) ensure zero hallucination.
• Mock Mode by Default: Generates "Golden" mock data matching the real Tunzaa API structure.
• Live Mode (Optional): Real-time verification against the Tunzaa Sandbox/Production.

🛠️ Usage (Live Mode)

To have the AI verify real data from your Tunzaa account (e.g., checking transaction statuses), add your credentials to the env block in your config:

"env": {
  "TUNZAA_API_KEY": "your_api_key",
  "TUNZAA_SECRET_KEY": "your_secret_key",
  "TUNZAA_ENVIRONMENT": "sandbox"
}

🏗️ Local Development

If you'd like to contribute or modify the server:

1. git clone https://github.com/Tunzaa/tunzaa_mcp.git
2. cd tunzaa_mcp && pnpm install && pnpm run build
3. Use the local path in your config: "args": ["/ABSOLUTE/PATH/TO/tunzaa_mcp/dist/index.js"]

License

ISC