This MCP server is a Java-based client wrapper that connects to an MCP Server using both stdio and HTTP Server-Sent Events (SSE). It wraps the mcp-client service and provides automated connection management, including automatic initialization at startup, periodic health checks, timeouts for idle connections, and connection reuse within a configured window. Once running, you can interact with the MCP Server either through standard input/output streams or via HTTP SSE endpoints exposed by the client, enabling real-time bidirectional communication with the server for prompts, responses, and streaming data. The included configuration supports isolation for multiple agents connected to the same MCP Server, ensuring that concurrent sessions do not interfere with each other while sharing underlying resources where appropriate.

Prerequisites:

• Java 17 or newer installed
• Maven or Gradle for building the project
• Access to the MCP Server and any required private model keys or endpoints

Installation steps:

1. Clone the repository: git clone <repository-url> cd <repository-directory>

2. Build the project (example with Maven): mvn -DskipTests package

   This will produce auto-mcp-client.jar in the target directory (path may vary depending on project setup)

3. Prepare configuration:

   • Ensure bootstrap.yml is configured with the correct NACOS config center and namespace
   • Import mcp-client-dev.yml into NACOS as described in the project docs
   • Update default-mcp-registry.json with your own API keys or MCP server entries
   • Ensure the node service path (if applicable) is configured so the stdio protocol can detect the node service availability

4. Run the server: java -jar target/auto-mcp-client.jar

5. Verify startup logs for successful MCP Server initialization and readiness for stdio or SSE interaction

Tips and common issues:

• Ensure JDK 17+ is used; some features rely on newer language/runtime features.
• If using NACOS for configuration, make sure the bootstrap.yml points to the correct config center and namespace, and that mcp-client-dev.yml is present and valid in NACOS.
• When using private OpenAI-compatible models, update the ai.openai configuration in your secret/config files with your API key and base URL as shown in the private model example.
• The system supports multiple agents connected to the same MCP Server with isolation between them. If you experience connection timeouts, consider adjusting idle timeout and health-check intervals in your configuration.
• For stdio interactions, ensure the process feeding stdin/stdout is properly synchronized with the MCP Server's messaging protocol to avoid deadlocks.
• Environment variables like OPENAI_API_KEY and OPENAI_BASE_URL should be kept confidential and injected securely in production.