Deep Agents, Batteries Included
A plain @Agent is a deep agent out of the box — long-term memory, a plan it maintains, a bounded virtual filesystem, and sub-agent delegation, all on by default via the harness. Narrow or opt down per agent. The same capability set as LangChain deepagents, hosted by a JVM framework. See the harness and Deep Agents vs LangChain.
Real-Time Core
@ManagedService with transport-agnostic Broadcaster pub/sub. WebSocket, SSE, long-polling, and gRPC — auto-negotiated per client. Rooms, presence, message history, and interceptors built in.
AI / LLM Platform
@Agent with pluggable AgentRuntime — auto-detects twelve runtimes: the built-in OpenAI-compatible client plus Spring AI, LangChain4j, Google ADK, Embabel, Koog, Semantic Kernel, AgentScope, Spring AI Alibaba, Anthropic, Cohere, and CrewAI adapters. Tool calling, conversation memory, filters, cost metering, and multi-model routing.
Multi-Agent Orchestration
@Coordinator + @Fleet for multi-agent systems. Parallel fan-out, sequential pipelines, and weighted routing. MCP, A2A, and AG-UI protocols for agent interoperability.
24-Chapter Tutorial
Build a complete real-time application from scratch. Covers transports, rooms, AI streaming, MCP, clustering, channels, and production deployment.
Spring Boot & Quarkus
First-class starters with auto-configuration, Micrometer metrics, and OpenTelemetry tracing. Tracks Spring Boot 4.0.7 and Quarkus 3.36.3. Under GraalVM Native Image, CI builds a native binary on each runtime and asserts one thing: a @ManagedService serving a real long-polling connection. WebSocket, SSE, transport fallback, and the AI stack under native are not covered by that lane. The reflection metadata ships inside atmosphere-runtime rather than in the starters, so it is not starter-specific — see Spring Boot and Quarkus.
