yu-ai-agent

The yu-ai-agent MCP server is a Java-based AI agent platform that exposes a set of MCP-compatible services enabling an autonomous AI persona to perform tasks such as multi-turn conversations, planning, tool invocation, and knowledge-grounded actions. It can integrate with an AI model provider and a knowledge base, call tools like web search, file operations, PDF generation, and image search, and expose these capabilities through MCP endpoints so clients can orchestrate complex AI workflows. Once running behind a container, you can connect to the MCP server to create agents, send prompts, and request MCP services (for example image-search MCP, document generation MCP, or plan-and-execute MCP flows) within your application or prototyping environment.

To use the server, ensure you have an MCP client or framework that supports the MCP protocol, then point the client to the server address exposed by the Docker container. The server will accept MCP service requests, route them to the appropriate tools or integrations, and return structured responses suitable for downstream processing or user-facing applications. Typical usage includes initiating an autonomous planning session, invoking a sequence of MCP services (search, fetch data, generate reports), and handling the agent’s memory/context across interactions.

Prerequisites:

• Docker installed and running on your machine
• Access to the yu-ai-agent MCP image (liyupi/yu-ai-agent) on Docker Hub or a private registry
• Optional: a compatible MCP client library or SDK to simplify interaction

Installation steps:

1. Pull and run the MCP image in Docker: docker pull liyupi/yu-ai-agent docker run -d --name yu-ai-agent -p 8080:8080 liyupi/yu-ai-agent

2. Verify the service is up by checking logs or hitting the MCP health endpoint (adjust port if needed): docker logs -f yu-ai-agent curl http://localhost:8080/health

3. (Optional) Configure environment variables for customization (these can be provided via docker run -e or a docker-compose file):

   • MCP_SERVER_PORT: Port on which the MCP server should listen
   • MCP_LOG_LEVEL: Logging level (e.g., INFO, DEBUG)
   • MCP_CONFIG_PATH: Path to a custom MCP configuration if the image supports it

4. If you prefer running locally without Docker, consult the project docs for alternative runtimes (e.g., Java/JVM run). The MCP server will still expose the same endpoints once started.

Tips and common issues:

• Ensure the Docker image liyupi/yu-ai-agent is accessible from your environment; you may need to login to Docker Hub for private registries.
• If you modify environment variables, restart the container to apply changes.
• When integrating with MCP clients, verify protocol compatibility and the endpoint paths defined by the server (health, endpoints for MCP services, etc.).
• If the server cannot reach external tools (web search, file systems, or PDFs), check network egress policies and required API keys for tools like SearchAPI or Pexels.
• For debugging, run in DEBUG mode if the image supports it and inspect agent memory/context handling and MCP call routing for complex workflows.