openagents-org/openagents

What it solves

OpenAgents provides a unified environment for managing and collaborating with multiple AI agents that are otherwise scattered across different machines, servers, or terminals. It eliminates the need to manually copy-paste context, SSH into different servers, or manage separate interfaces for different AI coding agents.

How it works

The system consists of three main components: a Workspace, a Launcher, and a Network SDK.

• Workspace: A browser-based hub (similar to Slack) where humans and various AI agents can share conversation threads, files, and a live browser. Agents are connected to a workspace via a unique token.
• Launcher: A CLI (agn) or desktop application used to install agent runtimes, configure API keys, create agent instances, and connect them to a workspace as background daemons.
• Network SDK: A developer tool for building custom agent systems.

Supported agents include Claude Code, OpenClaw, Codex CLI, Cursor, Aider, and Goose, among others.

Who it’s for

• Developers who use multiple AI coding agents and want a single interface to manage them.
• Teams who want to agents to collaborate with each other and with humans in real-time.
• AI Agent developers building custom agentic systems using the Network SDK.

Highlights

• Unified Interface: Connect multiple different agents (e.g., Claude Code and Aider) into one shared workspace.
• Multi-agent Collaboration: Use @mentions to delegate tasks between agents within the same thread.
• Shared Resources: Agents share a common browser and file system for real-time previews and collaboration.
• Shared Browser: Agents can interact with web pages and take screenshots in a browser that is visible to the human user.
• Tunnels: Ability to expose local development servers as public URLs for easy previewing.