Test recipient addresses
What replaces them
A simulator address is a way of saying “give me a bounce”. Lockally lets you say that directly, against a specific message, after the fact:
POST /v1/sandbox/messages/{id}/simulate
{ "event": "bounced", "reason": "550 5.1.1 unknown recipient" }
See Simulating failures for the full set.
This is better in two ways and worse in one.
Better: you choose the outcome per test, not per address. The same fixture recipient can deliver in one test and bounce in the next, with no second address to maintain and no mapping to remember.
Better: it cannot leak into production. A simulator address hardcoded in a fixture is an address, and addresses get copied into seed data and demo scripts and occasionally into live sends, where they bounce for real and land on your suppression list. There is nothing here to copy.
Worse: it is two calls, not one. You cannot assert “send to this address
and observe a bounce” in a single request. You send, read the id, then
simulate. For a test that only wants to exercise a webhook handler, that is one
extra line; for a load test that wants a stream of bounces, it is awkward.
What to use as the recipient itself
With a test key nothing is delivered, so the recipient address only has to be syntactically valid and not suppressed.
Use addresses in the reserved example domains, example.com, example.net and
example.org. They can never belong to a real person, which matters on the day
somebody runs the fixture with a live key by mistake.
Do not use @test.lockally.com. That suffix appears in the message_id of a
test send, and reusing it as a recipient makes log output ambiguous for no gain.
If you want delivery to a real inbox
Use a live key and an address you control. A catch-all on a domain you own is the usual arrangement, so each test can use a distinct address without creating mailboxes. That is a real send: it counts against your allowance and your reputation, and it belongs in a deliberate manual check rather than in CI. See Safe production testing.