The transport-agnostic real-time framework for the JVM.
WebSocket, SSE, Long-Polling, gRPC, MCP — one API, any transport.
Atmosphere was built on one idea: your application code shouldn't care how the client is connected. Write to a Broadcaster, and the framework delivers to every subscriber — whether they're on a WebSocket, an SSE stream, a long-polling loop, a gRPC channel, or an MCP session. The transport is pluggable and transparent.
The two core abstractions are Broadcaster (a named pub/sub channel) and AtmosphereResource (a single connection). Additional modules — rooms, AI/LLM streaming, clustering, observability — build on top of these.
jbang generator/AtmosphereInit.java --name my-app --handler ai-chat --ai builtin --tools
cd my-app && ./mvnw spring-boot:runGenerates a ready-to-run Spring Boot project with your choice of handler (chat, ai-chat, mcp-server), AI framework, and optional @AiTool methods. See generator/README.md for all options.
<dependency>
<groupId>org.atmosphere</groupId>
<artifactId>atmosphere-runtime</artifactId>
<version>4.0.10</version>
</dependency>@ManagedService(path = "/chat")
public class Chat {
@Ready
public void onReady(AtmosphereResource r) {
// r could be WebSocket, SSE, Long-Polling, gRPC, or MCP — doesn't matter
log.info("{} connected via {}", r.uuid(), r.transport());
}
@Message(encoders = JacksonEncoder.class, decoders = JacksonDecoder.class)
public ChatMessage onMessage(ChatMessage message) {
// Return value is broadcast to all subscribers
return message;
}
}What's New in 4.0 (full list)
Atmosphere applies the same philosophy to AI: your code shouldn't care which AI framework is on the classpath. Tools (@AiTool), conversation memory, guardrails, multi-backend routing, metrics, and observability are declared once with Atmosphere annotations and automatically bridged to Spring AI, LangChain4j, Google ADK, or Embabel at runtime. Per-endpoint model selection, auto-detected persistence (Redis/SQLite), and broadcast filter auto-registration round out the platform.
@AiEndpoint(path = "/ai/chat",
systemPrompt = "You are a helpful assistant",
conversationMemory = true,
tools = AssistantTools.class)
public class AiChat {
@Prompt
public void onPrompt(String message, StreamingSession session) {
session.stream(message); // auto-detects the AI framework from the classpath
}
}Tools are declared with @AiTool — framework-agnostic, portable across all backends:
public class AssistantTools {
@AiTool(name = "get_weather", description = "Get weather for a city")
public String getWeather(@Param("city") String city) {
return weatherService.lookup(city);
}
}Swap the AI backend by changing one Maven dependency — no tool code changes:
| Backend | Dependency | Bridged via |
|---|---|---|
| Built-in (Gemini/OpenAI/Ollama) | atmosphere-ai |
direct |
| Spring AI | atmosphere-spring-ai |
SpringAiToolBridge |
| LangChain4j | atmosphere-langchain4j |
LangChain4jToolBridge |
| Google ADK | atmosphere-adk |
AdkToolBridge |
| Embabel | atmosphere-embabel |
EmbabelAiSupport |
See spring-boot-ai-tools for the full tool-calling sample, spring-boot-ai-classroom for multi-persona conversation memory, and expo-client for React Native/Expo mobile chat. Four official framework samples have been forked and augmented with Atmosphere streaming: LangChain4j tools, Spring AI routing, Embabel horoscope, and ADK tools.
Already have a Claude Code, Copilot, Cursor, or Gemini CLI license? Embacle turns any CLI tool into an OpenAI-compatible LLM provider — no separate API key required.
LLM_BASE_URL=http://localhost:3000/v1 LLM_MODEL=copilot:claude-sonnet-4.6 LLM_API_KEY=not-needed \
./mvnw spring-boot:run -pl samples/spring-boot-ai-classroom@McpServer(name = "my-tools", path = "/atmosphere/mcp")
public class MyTools {
@McpTool(name = "ask_ai", description = "Ask AI and stream the answer to browsers")
public String askAi(
@McpParam(name = "question") String question,
@McpParam(name = "topic") String topic,
StreamingSession session) {
session.stream(question); // tokens broadcast to all clients on the topic
return "streaming to " + topic;
}
}| Module | Artifact | What it does |
|---|---|---|
| Runtime | atmosphere-runtime |
WebSocket, SSE, Long-Polling (Servlet 6.0+) |
| gRPC | atmosphere-grpc |
Bidirectional streaming transport (grpc-java 1.71) |
| Rooms | built into runtime | Room management with join/leave and presence |
| Module | Artifact | What it does |
|---|---|---|
| AI core | atmosphere-ai |
AiSupport SPI, @AiEndpoint, filters, routing, conversation memory |
| Spring AI | atmosphere-spring-ai |
Adapter for Spring AI ChatClient |
| LangChain4j | atmosphere-langchain4j |
Adapter for LangChain4j StreamingChatLanguageModel |
| Google ADK | atmosphere-adk |
Adapter for Google ADK Runner |
| Embabel | atmosphere-embabel |
Adapter for Embabel AgentPlatform |
| MCP server | atmosphere-mcp |
Model Context Protocol server over WebSocket |
| Module | Artifact | What it does |
|---|---|---|
| Redis | atmosphere-redis |
Cross-node broadcasting via Redis pub/sub |
| Kafka | atmosphere-kafka |
Cross-node broadcasting via Kafka |
| Durable sessions | atmosphere-durable-sessions |
Session persistence across restarts (SQLite / Redis) |
| Module | Artifact | What it does |
|---|---|---|
| Spring Boot | atmosphere-spring-boot-starter |
Auto-configuration for Spring Boot 4.0+ |
| Quarkus | atmosphere-quarkus-extension |
Build-time processing for Quarkus 3.21+ |
| Kotlin DSL | atmosphere-kotlin |
Builder API and coroutine extensions |
| atmosphere.js | atmosphere.js (npm) |
Browser & React Native client with React, Vue, Svelte, and RN hooks |
| wAsync | atmosphere-wasync |
Async Java client — WebSocket, SSE, long-polling, gRPC |
| Java | Spring Boot | Quarkus |
|---|---|---|
| 21+ | 4.0.2+ | 3.21+ |
JDK 21 virtual threads are used by default.
- Full documentation — architecture, configuration, and API reference for every module
- Project generator — generate a ready-to-run project with one command
- Samples — runnable apps covering every transport and integration
- Javadoc
Available via Async-IO.org
Apache 2.0 — @Copyright 2008-2026 Async-IO.org