case $? is safe to write against.
Code
10 is reserved and is never returned today.
Distinctions that matter in scripts
3 is never worth retrying. A rejected credential will not be accepted a moment later, and repeating it is how an address gets blocked. Fix the key instead. 9 and 124 answer different questions. 9 means the network failed. 124 means the time you allowed ran out, whether through--timeout or --exit-timeout. A job that retries on 9 and gives up on 124 is usually behaving correctly.
Two 403 responses map to 5. scheduling_not_included means the plan does not include scheduled sends. browser_not_allowed means the client configuration is wrong. Both are reported as preconditions rather than auth failures: the credential was valid, and the entitlement was not.
Nothing retries by itself. When the server asks for a wait, that is reported and the decision is left to whatever is running the command. A retry after 7 or 8 is only safe with an idempotency key.
Failure and stdout
On failure stdout is empty, so piping into a parser never yields half a document. In JSON mode an error is written to stderr in a consistent shape:mailkube doctor is the single exception. It exits 5 when a check failed, having written the whole report anyway: the code is a verdict on the report rather than a failure to produce one. That is what makes mailkube doctor && ./deploy.sh stop at a broken configuration while still telling you which check broke.
Looking up an error
550 5.6.0, so mailkube errors explain 550 works the same way:
Where each error is documented
The CLI’s catalog is a lookup for what you have in front of you. The authoritative descriptions live with the surface that produces them:- API errors for every name the REST API returns, with its status code and cause.
- SMTP error codes for what a submission server’s reply means.
- Rate limits for what triggers a 7 and how long to wait.