atmosphere

Stream Real Time and LLM response data over WebSocket, SSE, and the MCP protocol

Installation

Run this command in your terminal to add the MCP server to Claude Code.

Run in terminal:

Command

claude mcp add --transport stdio atmosphere-atmosphere java -jar atmosphere-mcp.jar \
  --env MCP_TLS="false" \
  --env MCP_HOST="localhost" \
  --env MCP_PORT="8080" \
  --env MCP_LOG_LEVEL="info"

How to use

Atmosphere includes an MCP server module that exposes a Model Context Protocol endpoint over WebSocket, allowing clients to connect and exchange commands and streaming data via MCP sessions. The MCP server is built on top of Atmosphere's transport-agnostic framework, so clients can connect using WebSocket or other transports (SSE, long-polling, gRPC) without changing the server logic. You can define MCP tools on the server side using annotated methods (e.g., @McpTool and @McpParam) and stream results back to connected MCP sessions. This enables real-time, streaming tool calls and responses to AI assistants or other agents that rely on MCP for interoperability.

How to install

Prerequisites:

Java 17+ and a JDK installation
Maven or Gradle for building from source (optional if using pre-built jars)
Git to clone the repository

Install and run from source:

Clone the repository git clone https://github.com/Atmosphere/atmosphere.git cd atmosphere
Build the MCP module (atmosphere-mcp) mvn -pl atmosphere-mcp -am clean package
Run the MCP server (example) java -jar atmosphere-mcp/target/atmosphere-mcp.jar
Verify the server is running by connecting a client to ws://localhost:8080 or the configured host/port

Notes:

If you prefer Docker, build a container image that runs the java -jar command and expose the MCP port.
You can adjust host/port with environment variables MCP_HOST and MCP_PORT as shown in the mcp_config example.

Additional notes

Tips and common considerations:

MCP sessions are transport-agnostic; ensure your client selects a compatible transport (WebSocket is the default).
Use MCP tool annotations (@McpTool, @McpParam) to expose server-side capabilities to AI agents or clients. Streaming tokens from long-running tools can be broadcast to topics or sessions in real time.
Configure logging (MCP_LOG_LEVEL) to troubleshoot connection issues or tool execution.
If running in a containerized environment, map the MCP port and ensure network policies allow WebSocket traffic.
For production, consider securing MCP with TLS and proper authentication/authorization on your MCP endpoints.

Related MCP Servers

ai-code-helper

489

2025 年 AI 编程助手实战项目（作者：程序员鱼皮），基于 Spring Boot 3.5 + Java 21 + LangChain4j + AI 构建智能编程学习与求职辅导机器人，覆盖 AI 大模型接入、LangChain4j 核心特性、流式对话、Prompt 工程、RAG 检索增强、向量数据库、Tool Calling 工具调用、MCP 模型上下文协议、Web 爬虫、安全防护、Vue.js 前端开发、SSE 服务端推送等企业级 AI 应用开发技术。帮助开发者掌握 AI 时代必备技能，熟悉 LangChain 框架，提升编程学习效率和求职竞争力，成为企业需要的 AI 全栈开发人才。

sonarqube

398

SonarQube MCP Server

quarkus-workshop-langchain4j

Quarkus Langchain4J Workshop

WigAI

Bitwig Controller Extension that provides an MCP Server for AI Agent control

obsidian

MCP server for Obsidian vault management - enables Claude and other AI assistants to read, write, search, and organize your notes

SchemaCrawler-AI

Free database schema discovery and comprehension tool