OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Agent Infrastructure

Atmosphere

An open-source Java runtime and real-time engine for building and serving AI agents on the JVM.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Atmosphere — OpenViST project profile in Agent Infrastructure.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Atmosphere when a JVM application needs stateful, streamed and governed agents integrated into its own hosting stack.

Who it is for

  • JVM application teams building AI agents
  • Developers needing portable AI runtime integrations
  • Teams requiring streaming, governance or human approval workflows
  • Organizations operating self-hosted, multi-channel agent services

Documented uses

  • Build an agent behind a common interface across supported AI runtimes.
  • Stream agent output to clients over documented real-time transports.
  • Gate tool calls with human approval and governance controls.
  • Resume longer-running sessions after reconnects or durable pauses.
  • Expose agents through MCP, A2A or AG-UI.
Before you choose

Requirements & limits

The repository is public and uses the Apache-2.0 license. It documents Java, server and self-hosted use, but no pricing or account requirement is provided.

  • Whether the application is built for the JVM and can provide its own host and scheduling.
  • Whether model-provider portability through the AgentRuntime SPI is important.
  • Whether the service needs WebSocket, SSE, long-polling, gRPC or optional WebTransport delivery.
  • Whether tool calls require policy controls, cost ceilings, PII rewriting or human approval.
  • Whether durable sessions, checkpoints and replay are needed for long-running work.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Run one agent across supported AI runtimes

Use a plain @Agent with Atmosphere’s common runtime interface across documented AI integrations.

Before you start

  • A Java/JVM application
  • Atmosphere runtime integrations
  • A supported AI runtime such as Spring AI, LangChain4j or Anthropic

Steps

  1. Define the agent as a plain @Agent class.

  2. Connect the application to Atmosphere’s AgentRuntime SPI.

  3. Select one of the documented runtime integrations.

  4. Run the agent through the shared interface.

Expected outcome

The documented adaptation is a portable agent implementation across supported runtime integrations.

Guide limitations

  • The evidence documents the common interface and integrations but does not provide complete dependency or execution instructions.
Documentation 1 ↗
Source-based practical guide

Stream agent output through a real-time transport

Route agent output through Atmosphere’s broadcaster to a client-facing transport.

Before you start

  • A JVM application using Atmosphere
  • An agent that produces output
  • A client connection using a documented transport

Steps

  1. Route agent output through Atmosphere’s broadcaster.

  2. Choose WebSocket, SSE, long-polling or gRPC for delivery.

  3. Use optional WebTransport over HTTP/3 only with the documented supporting setup.

Expected outcome

Agent output is delivered through one of the documented real-time transport paths.

Guide limitations

  • The evidence does not provide a complete client setup or deployment procedure for a specific transport.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
1 October 2026
Original website title
Atmosphere/atmosphere · Official homepage title · retained in its source language
How we review sourcesSuggest a correction