MCP connection issues

Troubleshoot problems connecting your AI tool to Contextium via MCP.

"Authentication failed" or stuck in a sign-in loop

Cause: Your browser session may have expired or you're signed into the wrong account.

Fix:

  1. Clear your browser cookies for contextium.io.
  2. Sign in again at contextium.io and confirm you can access your workspace.
  3. Restart your AI tool (Claude Desktop, VS Code, or Cursor).

"No workspaces found"

Cause: Your account isn't a member of any workspace, or you're signed into the wrong account.

Fix:

  1. Go to contextium.io/dashboard and confirm you see at least one workspace.
  2. If you were recently invited, check your email for the invitation and accept it.
  3. Restart your AI tool after confirming workspace access.

Tools don't appear in your AI assistant

Cause: The MCP server config may be incorrect or your AI tool hasn't loaded it yet.

Fix:

  1. Verify your config matches the format in Using MCP.
  2. Make sure you're using "type": "http" and "url": "https://mcp.contextium.io".
  3. Restart your AI tool completely (not just reload).

"Forbidden" or 403 errors

Cause: Your workspace may have exceeded its plan limits and the grace period has ended.

Fix:

  1. Check your usage at Settings > Usage in the Contextium web app.
  2. Upgrade your plan or reduce usage to restore access. See Understand usage limits.

Connection works but responses are slow

Cause: Large workspaces with many files may take longer to search.

Fix:

  1. Use specific search queries rather than broad ones.
  2. Check status.contextium.io for any ongoing performance issues.