MCPifex Errors: Causes and Fixes (Troubleshooting)
MCPifex errors come from a short, fixed list: 401, instance_disabled, connection_revoked, not_launchable and an expired idle session. Each one has a single cause and a quick fix. Find your symptom below and follow the fix.
Where you see the error
Errors appear in one of two places. Your AI app shows them when a request fails. The portal shows them when you click Test connection on an instance, which makes a real call before you finish setup. That's why you'll usually meet these errors there first. As of September 2026, these are the errors you can see for a bad key, an off instance or a lost connection.
Say Claude answers "the tool call failed" after you revoked a key an hour ago. Match the symptom to a section below.
I get a 401
Cause: the gateway didn't accept your API key. It's missing, mistyped, or revoked. A 401 shortly after you revoke a key is expected, since revocation takes effect within about a minute.
Fix: check that the key is sent in exactly one accepted form: an Authorization: Bearer mcpx_... header, or the key as the first path segment, /mcp/mcpx_.... If you still get a 401, generate a new key in the portal. To test a key outside your AI app, run the curl script in the quickstart. See API keys and authentication for both forms.
I get instance_disabled
Cause: the instance exists but is switched off. The usual reason is a plan downgrade. Instance count is the only hard limit, so MCPifex disables your newest instances beyond the new cap. Nothing is deleted.
Fix: move to a plan with enough instances on the pricing page. Upgrading re-enables disabled instances oldest-first. Instances and tools explains the limits.
I get connection_revoked
Cause: the instance uses a connected Google account, and that account's access no longer works. Usually someone removed MCPifex's access from their Google account.
Fix: open the instance in the portal and use Connect Google account again. The account must be an owner or full user of the Search Console property. The Google Search Console page has the details.
I get not_launchable
Cause: the server behind that instance isn't hosted by MCPifex yet, so there is nothing to launch.
Fix: use a hosted server instead. The marketplace shows what you can connect, and PostgreSQL is a good first one.
My session stopped after a while
Cause: MCPifex runs one isolated server process per client session. It stops that process when the session closes or sits idle. This is normal cleanup, not a fault.
Fix: nothing on your side. The client starts a new session on its next request. You don't need a new key or any portal action.
Quick reference
| Symptom | Cause | Fix |
|---|---|---|
401 | Missing, mistyped or revoked API key | Check the key, or generate a new one |
instance_disabled | Instance switched off, often after a downgrade | Upgrade the plan |
connection_revoked | Connected account lost access | Reconnect it in the instance |
not_launchable | Server isn't hosted yet | Use a hosted server |
| Session stops after idle time | Server process ended after inactivity | Start a new session, no action needed |
A tool is missing or refused
That isn't one of the codes above. The gateway only lists enabled tools and refuses calls to any other, so switch the tool on in the instance. Write tools start off. If nothing here matches, check the instance's call logs in the portal, or write to support@amdose.com.
Ready to connect?
Host any MCP server behind one endpoint and control exactly what your agents can reach.