MCP session still not preserving workspace/project between tool calls (Claude)

Ask us anything
1

Hi team — following up on the previous thread Project context not reachable through MCP, which was marked as resolved but I’m still hitting the same behavior on Claude.ai (web client), so opening a new topic as the old one is closed to further replies.

Reproduction (100% deterministic on my side, tested today from Claude.ai web):

  1. ping → Connected to WeWeb (userId=...). No workspace selected :white_check_mark:
  2. switchWorkspace(workspaceId=...) → Switched to workspace "..." :white_check_mark:
  3. switchProject(projectId=..., workspaceId=...) → Switched to project "..." :white_check_mark:
  4. Any next call — searchPages, listProjects, or even a bare ping with no arguments — → No active project on this MCP session / No workspace selected :cross_mark:

Key detail that may help narrow things down: even a plain ping immediately after a successful switchProject comes back with “No workspace selected”. So it’s not tied to read tools specifically — every subsequent call arrives at a fully reset session. The userId returned by ping stays consistent across calls, so auth itself is stable; it’s specifically the workspace/project selection that isn’t surviving between tool calls.

Already tried (same result each time):

  • Multiple back-to-back attempts within the same conversation
  • Fully disconnecting and reconnecting the WeWeb connector on the Claude side
  • Starting a fresh conversation
  • Passing workspaceId explicitly to every switchProject call

Was the previous thread resolved on the ChatGPT side only, or is a broader fix still in progress? Happy to share timestamps / userId privately for log lookup if useful.

Thanks!

2

Hey @gilsonricardomr :waving_hand:

Not sure why you’re not able to reply in the other thread. We didn’t close it on our side :confused:

In any case, I’ll pass on your message to the team! Feel free to also create a bug ticket here as well so you can share timestamps and userId privately.

3

Hey again!

Just got an update from the team: the issue should be fixed with the next release. ETA next week :slight_smile: