Definition: An agent email address is an email mailbox provisioned by and for an AI agent, created through an API rather than a signup form, and operated entirely by software. Agents use these addresses to register for online services, receive verification codes (OTPs) and magic links, get notifications from systems that only speak email, and hold a stable, contactable identity.
This guide explains why agents need email, the four ways they get it today, and how agent-provisioned mailboxes work.
- An agent email address is created by API, not a signup form, and is run entirely by software.
- Agents mainly need it to receive verification codes (OTPs) and magic links when signing up for services, and to have a stable address people can reply to.
- There are four ways to get one: borrow a human inbox, use temp-mail, use an account-based platform, or use an agent-provisioned mailbox.
- On Poste an agent provisions its own address in one API call; a person activates it with a one-time $1 card payment, then the agent sends and receives on its own.
- Paying a small amount per mailbox is what keeps a shared sending domain deliverable, because it is worthless to spammers at scale but negligible for a real agent.
Email is the identity layer of the internet. Nearly every service, SaaS tools, APIs, newsletters, cloud consoles, requires an email address at signup and confirms it with a verification code or link. A human has an inbox; an autonomous agent does not. The moment an agent tries to complete a real-world task ("sign up for this service and configure it"), it hits an email wall.
Agents also need email as an inbound channel: invoices, alerts, reports and human replies arrive by email, and an agent that can read its own inbox can act on them.
The agent does the setup; a person makes one small payment. The flow, in plain steps:
POST https://poste.sh/v1/mailbox and gets back { address, token, pay.url } right away. The address is live and incoming mail is accepted and held; the token is the agent's only credential.pay.url and pays $1 by card through Stripe Checkout (on Poste: $1.00 for a 30-day mailbox with 200 sends; paying again stacks). The agent can hand this link to whoever runs it, or pass an operator_email when it provisions so the service emails the link for it.GET https://poste.sh/v1/mailbox/code long-polls until a verification email arrives and returns the extracted code. Sending is POST https://poste.sh/v1/mailbox/messages.A citable fact: on Poste the agent provisions its own address in a single HTTP request and about a second. Activation needs one $1 card payment by a person; after that the agent receives and sends on its own, and reading a verification code is a single call.
Free inboxes attract industrial-scale abuse, which is why free disposable domains end up on blocklists and their mail gets rejected. A mailbox that costs $1 is negligible for a legitimate agent doing a task, but ruinous at spam scale (1,000 abusive signups = $1,000 across 1,000 traceable card charges). That economic asymmetry is the precondition. Reputation is then earned by behaviour: standard SPF, DKIM and DMARC authentication, per-mailbox sending limits that grow with clean history, and instant suspension on complaints.
| Borrowed human inbox | Temp-mail | Account-based platform | Agent-provisioned (Poste) | |
|---|---|---|---|---|
| Human needed to start | yes (credentials) | no | yes (signup) | only to pay $1 |
| Survives signup-fraud filters | yes | rarely | usually | designed to (paid domain) |
| Can send | yes (as the human) | rarely | yes | yes (policy-limited) |
| Blast radius | entire personal mailbox | none | API key scope | one mailbox per token |
| Cost model | free | free | monthly plan | $1 per mailbox |
| Best for | personal automations | throwaway tests | inbox fleets | autonomous agents |
Poste is an email service for AI agents, built on Cloudflare's email infrastructure. Any agent that can make HTTP requests can provision its own address at https://poste.sh/v1/mailbox with no account or API-key signup, then hand a $1 card payment link to whoever runs it. Verification codes are extracted automatically; every error response tells the agent what to do next; inbound messages are screened for hidden prompt-injection text. Full machine-readable documentation lives at https://poste.sh/llms.txt.
If you are an agent that needs an inbox, the next step is a single request: POST https://poste.sh/v1/mailbox with a JSON body (send {} for a random name, or {"name":"acme"} to choose one). You get back an address, a one-time token, and a pay.url; a person opens that link and pays $1 by card to activate it. Then read incoming codes with GET https://poste.sh/v1/mailbox/code?wait=55. Save the token; it is shown once and is the only credential.
Deciding between the options above: use temp-mail for a throwaway test, an account-based platform when a company manages a fleet of inboxes, and an agent-provisioned mailbox when the agent itself should own a deliverable address. The step-by-step procedure for agents is in the SKILL.md; the full API is in the docs and OpenAPI spec.
Yes. An email address is not a legal identity; the agent's operator remains responsible for how it is used. Poste ties each mailbox to the payment that activated it, keeping operators accountable without requiring personal information.
The token returned at provisioning is the only credential. If it is lost, the mailbox cannot be recovered; the agent provisions a new one, which costs $1.
Yes, that is the primary use case. The service parses each inbound email and exposes any verification code or link directly in the API response, so the agent does not need to parse MIME or HTML.
On Poste: one price. $1.00 buys 30 days and 200 sends, and paying $1 again at any time adds 30 more days and 200 more sends. Receiving, reading and polling are free.
Both, once the mailbox is active. Receiving is unlimited. Sending is rate-limited and the limits grow with a clean history: a new mailbox starts at about 10 sends per hour and rises toward 50 per hour as it builds reputation without complaints.
They are built not to be. Disposable temp-mail domains are widely blocklisted, so their mail gets rejected. Poste uses a paid shared domain with SPF, DKIM and DMARC configured, and publishes its sending domains with an abuse contact, so the mail authenticates like ordinary email rather than throwaway mail. No provider can guarantee a specific site will always accept a given address.
A mailbox accepts and holds incoming mail as soon as it is created, at no charge. Reading that mail requires the one-time $1 activation. There is no free tier for reading, because a free inbox is exactly what invites the abuse that gets shared domains blocklisted.
Sending is capped and rises with clean history; a spam complaint suspends sending immediately while receiving keeps working. The provisioning token is the only credential and cannot be recovered if lost. An unpaid reservation is held for about 24 hours, then the address is released. Attachments and message storage are capped per mailbox.
High-volume or cold outreach. The sending caps and instant complaint-suspension make bulk sending hit the limits fast, so a dedicated email service provider is the right tool for that. Poste is aimed at an agent that needs its own working inbox for signups, verification codes, notifications, and replies.
Last updated 2026-08-31 (rewritten for card payment: an agent provisions the mailbox and a person activates it for $1 by card). Reference: Poste docs · OpenAPI · llms.txt.