Atmosphere

Install the Atmosphere CLI using Homebrew on macOS/Linux, or via the curl installer for other platforms.

brew install Atmosphere/tap/atmosphere
# or
curl -fsSL https://raw.githubusercontent.com/Atmosphere/atmosphere/main/cli/install.sh | sh

Use the CLI to scaffold a new Spring Boot AI agent project with the ai-chat template. The --template flag selects from ai-chat, ai-tools, or multi-agent-startup-team.

atmosphere new my-agent --template ai-chat
cd my-agent

Pass your LLM provider API key as an environment variable when starting the application. The framework reads LLM_API_KEY by default, or adapter-specific keys like anthropic.api.key.

LLM_API_KEY=your-api-key ./mvnw spring-boot:run

For existing Spring Boot 4.0 projects, add the atmosphere-spring-boot-starter dependency to your pom.xml to enable the framework's agent capabilities.

<dependency>
  <groupId>org.atmosphere</groupId>
  <artifactId>atmosphere-spring-boot-starter</artifactId>
  <version>${atmosphere.version}</version>
</dependency>

Use Java annotations to define your agent class, prompt handler, and AI-callable tools. The @AiTool annotation makes a method available to the LLM for function calling.

@Agent(name="my-agent", description="A helpful AI agent")
public class MyAgent {

  @Prompt
  public Flux<String> chat(String message) {
    return runtime.stream(message);
  }

  @AiTool(description="Get current weather for a city")
  public String getWeather(String city) {
    return weatherService.fetch(city);
  }
}

Atmosphere exposes each agent as an MCP endpoint. Point your MCP client to the agent's MCP URL to access its tools and prompts over the standard protocol.

# MCP endpoint pattern:
# http://localhost:8080/atmosphere/agent/my-agent/mcp