Troubleshooting

Common issues and solutions when using the a4b.ai MCP server with your AI assistant.

Authentication Errors

Token expired or invalid

Your AI client shows an authentication error or stops responding to a4b.ai requests.

Solution: Most MCP clients refresh tokens automatically. If the error persists:

Remove and re-add the a4b.ai MCP server in your client
Re-authorize when prompted
Check if the application was revoked in Settings > Connected Apps

Insufficient permissions

Your AI client returns “Insufficient scope” when trying to create or update resources.

Solution: Re-authorize with both mcp_read and mcp_write scopes. Remove the server from your client config and add it again — you’ll be prompted to select scopes during authorization.

Not authorized for a resource

Your AI client returns “Not authorized” for a specific workspace or action.

Solution: Your user role doesn’t have permission for this action. Contact your organization admin to grant the appropriate role via Settings > Workspace Members in a4b.ai.

Data Errors

Resource not found

Your AI client can’t find an asset, workspace, or other resource.

Causes:

The resource ID doesn’t exist
The resource belongs to a different organization
The resource has been deleted

Solution: Ask your AI assistant to list resources first (e.g., “List my assets”) to find valid IDs.

Validation failed

Your AI client returns validation errors when creating or updating resources.

Solution: Check the required parameters and accepted formats for the tool. Ask your AI assistant “What parameters does [tool name] accept?” to see the tool’s schema.

Connection Issues

Client doesn’t show a4b.ai tools

Verify the server URL is exactly https://a4b.ai/mcp
Restart your MCP client after configuration changes
Check that your config file is valid JSON
Check client logs for connection errors

Check that your browser isn’t blocking popups from the MCP client
Ensure you have an active a4b.ai account

Token refresh fails

Refresh tokens are single-use — each refresh returns a new one
If your client used an old refresh token, re-authenticate through the full OAuth flow
Check if the application was revoked in Settings > Connected Apps

Plan Limits & Authorization

Plan limit reached

Your AI client returns an error when creating assets, workspaces, or inviting users.

Solution: Your organization has reached the capacity limit for the current plan. Check your usage with “Show my organization stats” — if you’re at capacity, upgrade your plan or remove unused resources.

PKCE failure (`invalid_grant`)

Your MCP client fails during the OAuth flow with invalid_grant or similar errors.

Causes:

The authorization code expired (codes are single-use and short-lived)
The PKCE code verifier doesn’t match the code challenge
The redirect URI doesn’t match what was registered

Solution: Remove and re-add the MCP server to start a fresh OAuth flow. Ensure your client supports PKCE (most modern MCP clients do).

Rate limited (429)

Your AI client receives “Too Many Requests” responses.

Solution: The MCP server rate-limits client registrations per IP. Wait a moment and try again.

Still stuck?

Ask your AI assistant “What’s my current user?” — this confirms your connection is working. If that fails, remove and re-add the server configuration.