# Workspace A **workspace** in AgentRQ is an isolated environment that groups together AI [agents](agent), [tasks](task), and team members. Each workspace has its own [MCP](mcp) endpoint, [task board](task-board), and configuration — providing a clean separation between different projects, teams, or deployment environments. ## What a Workspace Contains | Component | Description | |-----------|-------------| | MCP endpoint | The URL agents use to connect (`https://mcp.agentrq.com/{workspace_id}`) | | Task board | The human-facing UI for monitoring agent work | | Members | Team members who can view and respond to agent tasks | | Configuration | Custom context delivered to agents via `getWorkspace` | | Tasks | All tasks created by agents within this workspace | ## Workspace Isolation Each workspace is independently isolated: - Agents in one workspace cannot see or interact with tasks from another workspace - Each workspace has its own unique MCP URL and authentication - Members are scoped per workspace — you can share access with specific collaborators without exposing all workspaces ## Configuring a Workspace Workspace context is returned by the `getWorkspace` [MCP tool](mcp-tool). This gives the agent instructions specific to the workspace — like project context, coding conventions, or escalation policies. Agents should call `getWorkspace` when starting a new session to understand their current operating environment. ## Multiple Workspaces You might use multiple workspaces to separate: - **Environments** — production agents vs. development/staging agents - **Projects** — one workspace per major product or repository - **Teams** — isolate agent work by team or access level - **Clients** — separate workspaces for each client when doing consulting work ## Related Terms - [Task Board](task-board) - [Task](task) - [MCP Server](mcp-server) - [Channel](channel) - [Operator](operator)