JBang MCP Examples

The JBang MCP Examples server is a lightweight, Java‑centric showcase that demonstrates how the Model Context Protocol (MCP) can be leveraged to embed executable tools directly into AI assistants. It addresses the common challenge of bridging static codebases with dynamic conversational agents, enabling developers to expose reusable Java utilities as first‑class MCP services without the overhead of building full microservices from scratch.

At its core, the server bundles a set of small Java applications—such as a calculator, real‑time weather fetcher, and stopwatch—that are compiled and run through the JBang CLI. Each tool is packaged as an MCP endpoint, exposing its functionality via a JSON‑based request/response schema. This makes it trivial for an AI assistant to invoke the calculator, retrieve live weather data, or start a stopwatch—all while preserving type safety and leveraging Java’s ecosystem (Maven, Docker, Quarkus). The integration is seamless: the MCP server listens on a configurable port (e.g., ), and any AI client that understands MCP can discover and call these tools on demand.

Key capabilities include:

• Tool Exposure: Each Java class is wrapped as an MCP tool, automatically generating the necessary metadata for discovery.
• Containerization: The examples illustrate Docker and Docker Compose workflows, enabling the tools to run in isolated environments or as part of a larger stack.
• Dependency Management: By leveraging Maven’s version‑display plugins, developers can keep the toolchain up to date without manual intervention.
• Cross‑Platform Invocation: The JBang CLI allows the same tool to be executed locally, from CI pipelines, or within Docker containers, making it flexible for various deployment scenarios.

Real‑world use cases span from rapid prototyping—where a data scientist can ask an AI assistant to compute a complex formula—to production workflows, where the weather tool could be integrated into a smart home assistant. Developers benefit from reduced boilerplate: they write pure Java logic once, expose it via MCP, and immediately gain an AI‑ready interface. The server’s minimal footprint and reliance on standard Java tooling mean it can be dropped into existing projects with zero friction, while still providing the full power of MCP’s resource and sampling abstractions.