This MCP server is a Spring Boot-based MCP server that exposes a JSON-RPC over HTTP/SSE interface for interacting with AI tooling. Clients connect to the SSE endpoint to establish a session and then send JSON-RPC requests to the message endpoint to initialize and retrieve available tools, and to invoke specific tools. For example, you can connect to the SSE endpoint at /sse to establish a session, receive a data message containing the sessionId, and then POST to /mcp/message?sessionId={sessionId} with methods like notifications/initialized or tools/list to discover available tools. In the sample wiring, tools such as add (for arithmetic addition) and owner_getByUserName (fetching data by name) are exposed, each with its own input schema. The flow is: (1) open SSE connection to /sse, (2) request tool list via a tools/list or initialized notification, (3) use the sessionId shown in the data payload to POST tool invocations and receive results, all via JSON-RPC payloads.

Prerequisites:

• Java 17 (or compatible JDK) installed
• Maven installed (optional if using a pre-built jar)
• Internet access for dependency resolution

Installation steps:

1. Clone the repository or obtain the project sources.
2. Build the project (if you have source): mvn -version # ensure Maven is installed mvn clean package

   after build, the jar is typically located under target/ (e.g., target/mcp-spring-ai-0.0.1-SNAPSHOT.jar)

3. Run the MCP server: java -jar target/mcp-spring-ai-0.0.1-SNAPSHOT.jar
4. Verify the server is running, then connect to the SSE endpoint at http://localhost:8090/sse and follow with tool discovery at http://localhost:8090/mcp/message?sessionId=YOUR_SESSION_ID.

Notes:

• If you run via an IDE, ensure the proper Maven dependencies are resolved and the Spring Boot app starts with the embedded container.
• If you want to run from a pre-built jar, adjust the path in mcp_config accordingly.

Tips and common issues:

• Ensure Java 17 is used; the project targets Java 17.
• The SSE and message endpoints are typically served under /sse and /mcp/message respectively unless customized via configuration (see yml parameters in the repository for spring.ai.mcp.server.* properties).
• If the tools list is empty or tool invocations fail, verify that the server has initialized correctly (notifications/initialized) and that the sessionId used in subsequent requests matches the one provided in the SSE data payload.
• Check environment variables and Spring profiles if environment-specific behavior is needed; you can enable verbose logging to troubleshoot transport or JSON parsing issues.
• When using the Maven build, ensure network access to fetch Spring AI BOMs and related dependencies; corporate networks may require proxy configuration.