Sindi AI MCP Server

Overview of the Sindi AI MCP Server

The Sindi AI MCP Server is a Java‑based implementation of Anthropic’s Model Context Protocol (MCP), designed to bridge AI assistants with Jakarta EE applications. By exposing MCP resources, tools, prompts, and sampling capabilities through either REST or Servlet endpoints, it solves the common pain point of integrating sophisticated AI services into existing enterprise stacks without reinventing protocol handling or transport layers. Developers can now expose custom tool functions, prompt templates, and data resources directly from their application code, enabling AI assistants to call back into business logic or external APIs in a type‑safe and declarative manner.

At its core, the server is modular: the SPI module defines CDI‑friendly annotations and interfaces; the runtime modules provide concrete implementations for servlet or REST transports, including SSE support for streaming responses; and the features module demonstrates how to register real‑world capabilities. This separation allows teams to pick only the parts they need, keeping deployments lightweight while still offering full MCP compliance. The use of Jakarta EE standards (CDI, REST, Servlet) ensures seamless integration with familiar tools like Maven, WildFly/LB, or Liberty.

Key capabilities include:

• Tools: Annotate any CDI bean method with or LangChain4J equivalents. The method must return a string, and the server automatically exposes it as an MCP tool callable by an AI assistant.
• Prompts: Use and to expose prompt templates. Methods can return single or multiple objects, allowing dynamic conversation contexts.
• Resources & Templates: Mark methods with or to serve static or templated content. The return types (, ) give fine control over binary or text payloads.
• Programmatic registration: For advanced use cases, the , , and beans let developers register callbacks at runtime, offering flexibility beyond annotation‑based discovery.
• Logging & context: Inject for per‑session logging and to access session state or manually register features during a request.

Real‑world scenarios that benefit from this server include:

• Enterprise automation: A finance application can expose a tool that queries account balances, which an AI assistant then uses to answer user questions or generate reports.
• Dynamic content generation: A CMS can serve prompt templates that pull in current article metadata, enabling an assistant to draft or edit content on the fly.
• Secure data access: By tying resource endpoints to authenticated sessions, developers can expose sensitive documents or configuration files only to authorized AI interactions.
• Hybrid workflows: Combining REST and Servlet transports lets teams run the server in legacy servlet containers while still leveraging modern SSE streams for real‑time updates.

Because the Sindi AI MCP Server adheres strictly to MCP specifications and leverages Jakarta EE’s robust ecosystem, it offers a production‑ready, scalable solution for developers looking to embed AI assistants into their applications without managing low‑level protocol details.