How to Use Workspaces in OpenClaw?

Workspaces let you run multiple independent AI agents on a single OpenClaw installation. Each workspace has its own soul.md, memory, skills, and even a different AI model — completely isolated from other workspaces.

Think of workspaces as separate AI personalities on one server. You might have: - A personal assistant workspace (friendly, casual, connected to personal calendar) - A work agent workspace (professional, connected to Slack and GitHub) - A trading bot workspace (analytical, connected to exchange APIs)

Create a workspace with the CLI command. Each workspace gets its own directory under ~/.openclaw/workspaces/ with independent configuration.

Workspaces can connect to different messengers. Your personal assistant might be on Telegram while your work agent is on Slack. Or you can connect the same messenger to multiple workspaces using different bot accounts.

Resource usage: each workspace runs its own MCP skill servers, so more workspaces mean more memory usage. A typical workspace uses 200-500 MB RAM depending on installed skills.

Switch between workspaces in the CLI or TUI. The Dashboard shows all workspaces with their status and activity.

Tip: Start with a single default workspace. Create additional workspaces only when you need genuinely different agent personalities or isolated skill sets.

# Create a workspace
openclaw workspace create work-agent

# Switch to workspace
openclaw workspace use work-agent

# List all workspaces
openclaw workspace list