Testing mailbox

Plenty of flows only finish in an inbox: a signup that waits for a verification code, a magic-link login, a notification your app expects someone to answer. Agentiqa gives every run its own email address, and the agent both reads it and replies from it — no mailbox of yours involved, nothing for you to set up.

What you get

Each run receives a unique address, for example a-agent+m9x1y2-3z4w5v@agentiqa.com. It is real: mail sent to it is delivered, and the agent can read it during that run. A new run gets a new address, so a signup test can register fresh every time instead of colliding with the account the last run created.

Two things the agent can do with it:

• Read — open the verification email, take the code or the confirmation link, and finish the flow.
• Reply — answer a message your app sent, from the same address your app wrote to, threaded into the same conversation.

You do not configure any of it. Write the test in plain language and the agent uses the address where it is needed.

Signup with a verification code

Sign up with a new account and complete the email verification.

The agent fills in the form with the run's address, submits, opens the inbox, takes the code out of the message, and enters it. If the email is slow, it waits and checks again instead of guessing; if it never arrives, it says so rather than claiming the account was verified.

In a saved test plan the address is written as {{unique-email}}:

Open the signup page
Enter {{unique-email}} in the Email field
Submit the registration form
Verify the account is confirmed

Every replay resolves that token to a fresh address, so the plan keeps working on the hundredth run.

Approve by email

Request approval on this page, then approve it by replying to the email with APPROVED.

Your app emails the run's address; the agent reads the message and answers it with the text your test specifies. Your app sees a normal reply — same sender address it wrote to, correct threading — so whatever it does with inbound mail happens for real.

Two properties are worth knowing, because they shape what you can ask for:

• The agent can only answer mail that actually arrived at its address. It cannot email an arbitrary person, and there is no way to ask it to.
• It replies with what your test says. If you want an exact payload — a keyword, a reference number, a decision word — put it in the instruction.

What the agent will not do

• Invent a verification code, a link, or the contents of an email it has not read.
• Claim a reply was sent when delivery failed — it reports the failure instead.
• Use a throwaway mailbox service if you ask for one; it uses the built-in address, which is the one it can actually read, and tells you it did.

Using your own address instead

If you give the agent a specific email address in the conversation, that identity wins: the agent uses it for the flow and treats the inbox as yours. It will not read or send mail on your behalf — when a verification code is needed, it pauses and asks you for it.