The agent's address — tinysend agents
Most agent tasks stall at “we sent a code to your email”. This is the surface that gets past it, without handing an agent somebody’s inbox.
What the agent gets
- one address: the person’s tinysend name, for example
anton@tinysend.com - a label per site: put a dot and any word after the name.
anton.thatshop@tinysend.comarrives in the same inbox, and nothing has to be created first — the agent types it into the form and mail comes - a wait tool that returns the code or the confirmation link in the mail that follows
Dot, not plus: forms accept a dot, and list cleaners do not strip it. + keeps its existing meaning, the action suffix.
What it cannot do
- read mail sent to the address itself — only mail that arrived through a label it made
- read a password reset, a sign-in alert, account recovery, or a two-factor change. Those are delivered to the person and withheld here
- see newsletters, subscribers, contacts, or any other mailbox
- send anything
That is the agent scope, and a token holding it gets a server with only these tools on it.
The tools
get_agent_address— the address to build labels fromwait_for_email— wait for mail at a label and read the code or link out of itlist_agent_mail,get_agent_mail— what has arrivedlist_agent_aliases,create_agent_alias— the labels in use
wait_for_email
Call it straight after submitting the form.
{
"address": "anton.thatshop@tinysend.com",
"timeout_seconds": 90,
"from_contains": "thatshop"
}
It answers as soon as mail arrives:
{
"email_id": "01k5...",
"from": "noreply@thatshop.example",
"code": "482913",
"code_confidence": "high",
"code_evidence": "code is 482913",
"link": null
}
timeout_secondsdefaults to 60, maximum 180, and0checks once and returnssincedefaults to now, so mail already sitting in the box is ignored- a code is never guessed:
code_confidenceishighwhen a label word sat next to it,lowwhen it was a lone number in an email about verifying, and the result isnullwhen neither was true - when the mail was a reset or an alert,
withheldexplains that and there is no code
For the person
Every label an agent invents shows up in the account under the address, with who first wrote to it and how many messages it has taken, and a switch that stops it. Switching a label off refuses its mail at the door; the rest of the address is untouched.