MCP Alchemy Server

You can access payout and transaction data from the XB database through an MCP server designed for use with MCP clients. It enables payout transaction lookups by UUID or Client Transaction ID and is built to be compatible with any MCP client.

How to use

Connect your MCP client to the server URL and use the provided payout transaction lookup tool to retrieve details from the XB database. You can search by a UUID or by a Client Transaction ID. When searching by Client Transaction ID, set the appropriate flag so the server treats the input as a client id. The server adheres to the MCP standard, ensuring broad compatibility with supported clients.

The available operation lets you fetch a payout transaction, which returns the relevant transaction data from XB. Use this to verify payouts, reconcile records, or feed transaction details into downstream processes via your MCP client workflow.

How to install

Prerequisites you need before installation:

• Python 3.12 or higher

• uv (recommended for dependency management)

Follow these concrete steps to get the server running locally:

git clone <repository-url>
cd oncall-mcp
uv sync

Additional configuration and running notes

Configure environment variables to enable database access. Copy the example environment file, then edit it to include your database connection strings.

cp .env.example .env

Start the server for development

To run the server locally, use the development command with hot-reloading if your runner supports it.

uv run main.py

Configuring MCP clients to connect

Your MCP client needs to know how to reach this server. An example client configuration indicates the MCP server URL that points to the local development instance.

{
  "mcpServers": {
    "PivotOnCall": {
        "serverUrl": "http://localhost:8000/mcp"
    }
  }
}

Available tools

get_payout_transaction

Retrieves a payout transaction from the XB database using either a UUID or a Client Transaction ID. You can specify by_client_id to search by Client Transaction ID; otherwise, the search uses a UUID by default.