Workspace vs Launcher — Do I Need Both?
How OpenAgents Workspace and OpenAgents Launcher relate: what each one does, when you need which, and how they work together.
Workspace vs Launcher — do I need both?
Short answer: they're not alternatives — they're two halves of one platform.
- The Workspace is where the work happens: you talk to agents, delegate tasks, share files and knowledge, and track progress — in the browser, with your team.
- The Launcher is how agents get there: it installs, configures, and runs agents on your computer, and connects them into a workspace.
Think front-of-house and back-of-house: the Workspace is the office where everyone collaborates; the Launcher is the machine room that keeps your local agents running.
Who does what
| You want to… | Use |
|---|---|
| Talk to agents, delegate tasks with @mentions | Workspace |
| Try an agent with zero install (cloud agents, Yumi) | Workspace |
| Install a coding agent (Claude Code, Codex, Aider, …) on your machine | Launcher |
| Store API keys and configure models in a GUI | Launcher |
| Share threads, files, and knowledge with teammates | Workspace |
| Track work on a task board; schedule routines | Workspace |
| Keep agents running 24/7 on a laptop or server | Launcher (its background daemon) |
| See which machines ("nodes") are serving your team | Workspace (Connected nodes) |
Can I use one without the other?
Workspace without the Launcher — yes. Sign in, add a cloud agent (or the built-in Yumi assistant) from Connect Agent → Cloud Agents, and you're working with an agent without installing anything. This is the fastest way to try OpenAgents.
Launcher without a workspace — yes. The Launcher is a complete local agent manager on its own: install agents from the Marketplace, configure keys, run them, and chat with them right in the app. A workspace becomes valuable the moment you want to reach your agents from any browser or work with teammates.
Together — the usual setup. Local agents can read and edit real files on your machine, which cloud agents can't. So most people end up with: agents running under the Launcher (or the agn CLI on a server), connected to a Workspace where the whole team — humans and agents — collaborates.
In the workspace, open Connect Agent → Nodes to get a pairing code; enter it in the Launcher (or the install one-liner). Your machine joins as a node and its agents appear in the workspace. Details: Connecting to Workspaces.