New MCP tool: list_workspace_members

You can now ask any MCP-connected AI client (Claude Desktop, Cursor, Claude Code, etc.) about your workspace teammates and get an answer. Previously, asking "who's on my team?" or "find Alice on my team" routed agents to search_people — which searches your customers, not your team — and returned nothing useful, even when the person was sitting right there as a workspace member.

What's new

A single tool: list_workspace_members.

• No args → returns everyone with login access to the active workspace, with their role (owner/admin/member) and the caller marked as isYou
• Optional query arg → filters by name or email substring

That's it. One tool, two behaviors, picked between by whether you pass an argument.

Why it's a separate tool from search_people

In Simple Product:

• People = your customer / CRM contacts. Created automatically when your SDK fires identify(). Tracked via the people table. This is what search_people searches.
• Workspace members = the humans with login access to your Simple Product workspace. Your team. Tracked via the workspaceMembers table. This is what list_workspace_members returns.

These are two genuinely different tables for two genuinely different concepts. We didn't want to overload one tool to do both — agents would have to guess which one you meant, and would be wrong half the time. So they're split, with descriptions that explicitly contrast.

How to use it

In any MCP-connected AI client:

"Who's on my team?"

"Find Alice on my team."

"List the admins of this workspace."

"Show me workspace members whose email is at acme.com."

The agent will pick list_workspace_members (with or without the query arg). Your team comes back with names, emails, and roles.

How this got here

Arthur Torres at Stable Kernel hit the gap and reported it as a bug — they were trying to look up themselves as a workspace member via search_people, came up empty, and very reasonably assumed the search was broken. It wasn't broken; we just hadn't exposed the right tool. Thanks Arthur — this one's yours.

Architectural footnote (for the curious)

This tool shipped with zero changes to the npm package — we added it server-side to /api/mcp/v1 and every existing 0.2.x bridge picked it up on its next session. No npm install, no version bump, no nag to upgrade. That's the model we settled on with the MCP V1 work last week, and this is the first time we used it in response to specific customer feedback — feedback to fix to ship in a single afternoon, no friction for anyone running the SDK.

If you're already on @simple-product/mcp@0.2.x, just restart your AI client and ask it about your team.