I'm the organization admin (and only user) of my Atlassian organization, and I'm trying to authorize an OAuth client ("Muse") for Trello's official MCP server (https://mcp.trello.com/v1). At the workspace consent step, selecting my Trello workspace shows "Permissions are blocked for these workspaces — Contact your organization admin to request access" and the Accept button stays disabled. The same thing happens for both of my workspaces.
Everything support documentation and the support AI pointed to has been checked:
- Atlassian MCP server settings → Domains: the client's callback domain (https://agent.meta.ai/api/hatch/oauth/callback) is allowlisted. I also removed and re-added it and saved, with no change.
- MCP server → Permissions: Read 16/16, Write 16/16, Search 4/4 all allowed, including the read_trello and write_trello toolsets. (Delete and Manage are blocked, which I understand is expected.)
- Data security policy: the "Atlassian MCP server" control is set to Allowed with 0 overrides.
- I captured the consent URL and decoded it: the request sent only the scopes "offline_access read:board:trello write:board:trello" for resource https://mcp.trello.com/v1 — no account-level Inbox/Planner scopes and no organization scopes.
- Update: I have since verified and claimed my domain in Atlassian Admin, so my account is now a claimed, managed account (this was not the case during the earlier checks). A fresh consent retry after that, signed out and back in (~06:22 UTC, Oct 30), still shows the same blocked banner.
So every documented control is set to allow, and the block persists. It looks like the back-end organization/workspace permission evaluation disagrees with what the admin UI shows. Has anyone else hit this with the Trello MCP server, or can an Atlassian team member check the evaluation for my organization? I can provide the organization ID, the full consent URL, and screenshots (blocked consent screen, permissions pages, data policy) privately.
Note: I've tried twice to file this through the support contact form as well; the form itself failed both times at the submission stage, which is why I'm posting here.