Spring AI MCP Batch Job Server

What is Spring AI MCP Batch Job Server?

A Spring Boot Model Context Protocol (MCP) server that exposes batch processing tools for financial transactions. It integrates Spring AI’s MCP Server Boot Starter (STDIO transport) with Spring Batch and OpenAI to read CSV files, categorize transactions by type, and persist results in PostgreSQL. Designed for AI assistant workflows (e.g., Claude Desktop) that need to launch, monitor, or retry batch jobs.

How to use Spring AI MCP Batch Job Server?

Build the project with ./mvnw clean install to produce a JAR. Then configure an MCP client (e.g., Claude Desktop) with a JSON entry specifying the java command, the JAR path, and the OPENAI_API_KEY environment variable. No web server is started; the client launches the server automatically over STDIO.

Key features of Spring AI MCP Batch Job Server

• Three MCP tools: startJob, restartJob, getJobStatus
• Reads CSV files from the user's Downloads directory
• AI-powered transaction categorization via OpenAI (gpt‑4o)
• Spring Batch pipeline with file movement, processing, and cleanup steps
• Persists results to PostgreSQL tables automatically
• Full STDIO transport; no HTTP server required

Use cases of Spring AI MCP Batch Job Server

• An AI assistant triggers a batch job to process a daily CSV of expenses.
• A user restarts a failed job by providing the execution ID.
• A monitoring workflow repeatedly checks job status until completion.

FAQ from Spring AI MCP Batch Job Server

What runtime dependencies does Spring AI MCP Batch Job Server require?

Java 21+, a running PostgreSQL instance, and an OpenAI API key set as the OPENAI_API_KEY environment variable.

How does the batch processing pipeline work?

It moves the input CSV to a safe location, reads and categorizes transactions using AI, writes data to PostgreSQL tables, then cleans up temporary files.

What transport does the server use?

STDIO transport only. The server has spring.main.web-application-type=none and is intended to be launched automatically by an MCP client.

Where does the processed transaction data get stored?

In a PostgreSQL database with tables: financial_transactions, credit_transactions, credit_card_payments, and categories. Spring Batch meta‑data tables are also created automatically.

How do I configure the server for Claude Desktop?

Add a JSON entry under mcpServers with the java command, the JAR path, and the OPENAI_API_KEY environment variable as shown in the README.