Email infrastructure for AI agents
anymail gives each AI agent its own real mailbox, so it can send and receive ordinary email on behalf of the person it works for: replies, confirmations, scheduling, and support conversations.
Get in touchRead our Acceptable Use PolicyFeatures
A standards-based mail service built for agents that act like people: one identity, one inbox, normal volumes.
One mailbox per agent
Every agent gets its own address and inbox, so conversations, replies and history stay with the agent that owns them.
Standard protocols
IMAP, SMTP submission and JMAP. Agents and their owners can use any standards-compliant client or library.
Authenticated by default
Outgoing mail is signed with DKIM and covered by SPF and DMARC, so recipients can verify where it came from.
Encrypted transport
Client connections require TLS, and server-to-server delivery uses TLS whenever the receiving server supports it.
Built-in limits
Sending is rate-limited per mailbox. anymail is meant for conversational mail, not bulk or marketing campaigns.
Human accountability
Every mailbox belongs to a person or team who is responsible for what their agent sends, under our Acceptable Use Policy.
How it works
anymail behaves like an ordinary email provider. The difference is that the mailbox user is an agent.
- Create a mailbox for your agentThe account owner sets up an address for an agent and keeps control of its credentials.
- Connect the agentThe agent signs in over IMAP/SMTP or JMAP with its own credentials, just like a regular mail client.
- Send and receive normallyThe agent reads incoming mail and sends replies and messages to people who expect to hear from it.
- Stay in controlThe owner can review the mailbox, change credentials, or disable the agent at any time.
No spam, ever
See the Acceptable Use Policy for details. To report abuse involving an anymail address, write to abuse@anymail.si and include the full message headers.