Studio MCP: Multi-Agent Improvements and Connected AI Clients
Key Takeaways
Every MCP tool now takes a studio_id, so agents address Studios explicitly instead of relying on session state. Multi-Studio, multi-client workflows are more reliable, clients pick up the change automatically, Assistant Settings shows connected AI clients, and we fixed two stability issues.
Hi Creators,
We’re making a round of updates to the Studio MCP Server that lets you connect MCP-enabled AI agents to Studio and make changes directly to Studio sessions.
Multi-Agent Improvements
Before this release, an MCP client was bound to a single Studio instance through the setactivestudio tool. If several agents shared a single MCP connection, they couldn’t reliably address different Studios.
Routing is now explicit. Each tool call must specify the Studio it targets instead of relying on session state. This matches how stateless MCP servers are meant to work, and it makes workflows with multiple Studio instances and multiple clients more reliable.
To get this update, restart your AI client once after Studio updates itself. From there, clients like Claude Code and Codex pick up the new tools on their own.
Key updates in this version include:
• studioid on every tool call: Each call names the Studio instance it targets. setactivestudio is removed, so if you were calling it directly, pass studioid instead.
• Place ID in listrobloxstudios: The tool call also returns the Place ID alongside its name and Studio instance ID, with the exception of local places that don’t have one, so those are still listed by name only. The Place ID lets an agent pick the right Studio when two open Studios share a name, to further improve multi-agent support.
Connected AI Clients
Assistant Settings now shows which AI clients are connected to the Studio MCP Server. Hover the tooltip on the MCP Servers page to see the list. If an agent isn’t responding, check here to confirm its MCP connection is live.…