How can my coding agent switch between workspaces in Rius?

Rius has no separate switch command. Your coding agent switches workspaces through the Rius MCP server by calling list_workspaces to find them and passing a workspace name or ID to the trace, metrics and alert tools. When it leaves the argument out, the tools use your default workspace. The reply comes back as Markdown.

How it works

Two kinds of tools are involved:

• Finding workspaces. list_workspaces takes no arguments and returns the id and name of each workspace you can access. On an OAuth sign-in it spans every organization you belong to, grouped under each organization’s name. get_me also lists your organizations, your role in each, and their workspaces.
• Choosing one. The trace, metrics and alert tools take an optional workspace argument. Examples are list_agent_traces, list_spans, agent_traces_summary, workspace_metrics_overview and list_alerts. Pass a workspace name or ID, or leave it out to use your default workspace.

If the same workspace name exists in more than one of your organizations, pass it as org-name/workspace-name. An ambiguous name returns an error.

Which workspaces your agent can reach depends on how it signs in:

• OAuth session. The tools see exactly the workspaces your user can access.
• Read-scoped API key. The key is bound to the single workspace it was created in. Other workspaces report as not found, so a headless agent does not switch at all. Give it one key per workspace.

Nothing the MCP server offers can create, rename or delete workspaces, or change members. Those stay in the console.

Set it up

1. Connect your agent to the Rius MCP server if you have not yet. If you use the Rius plugin, the server is already registered as rius. For Claude Code without the plugin:

   claude mcp add --transport http rius https://mcp.eu.console.rius-glassflow.com/mcp

   The MCP page has the steps for other clients and for headless keys.

2. Ask in plain words which workspaces you have:

   Which Rius workspaces do I have?

   Your agent calls list_workspaces with no arguments:

   { "name": "list_workspaces", "arguments": {} }

3. Name the workspace in your next question. For example:

   How did my agents do over the last 24 hours in the staging workspace?

   Your agent passes the name in the workspace argument. Here acme/staging is your own organization and workspace, written in the org-name/workspace-name form:

   {
       "name": "agent_traces_summary",
       "arguments": { "hours": 24, "workspace": "acme/staging" }
   }

4. To go back to your default workspace, ask again without naming one. Say “in the default workspace” if you want to be explicit, and the agent leaves workspace out.

The tool reference lists every tool and its arguments. If a question does not pick the tool you expect, name the workspace and the kind of data you want, such as traces, alerts or the metrics overview.

What you get

• A list of workspaces. list_workspaces returns each workspace with its id and name, grouped by organization when you belong to several. Use those values for the workspace argument.
• Your identity and access. get_me returns your email, your organizations with your role in each, and the workspaces you can reach.
• Answers scoped to the workspace you name. Trace lists, span waterfalls, cost and token summaries, alerts and the metrics overview all come back as Markdown tables and summaries that your client renders.
• Clear errors. An ambiguous workspace name returns an error, and the org-name/workspace-name form resolves it. A key-authenticated agent asking for another workspace gets not found.

For what to ask once you are in the right workspace, see Monitor AI agents in production.

When something else fits

• If you need to create, rename or delete a workspace, or change its members, use the Rius console. The MCP server cannot do it.
• If a headless agent must read several workspaces, mint a separate read-scoped key for each, or use an interactive OAuth session that spans all your workspaces.

Related