Guides

How email works for AI agents and humans, and how to keep both safe.

Agent inboxes

  • What is an agent inbox?

    An agent inbox is an email address an AI agent owns and works through an API. How it differs from a human mailbox and a sending API, and what to check.

  • Email for AI agents

    Why AI agents need email, the five ways to give an agent an address, what can go wrong when a model reads mail, and what to look for before you choose.

  • How to give an AI agent an email address

    A hands-on guide: get an address for your AI agent with one call, wait for mail, read it safely and reply, with working Python and TypeScript examples.

  • What an email API for AI agents must do

    What an email API for AI agents does that a sending API does not: receive and store mail, keep threads, wait for a message, extract codes and screen.

  • Agent identity and email: why an agent needs its own address

    Why an AI agent should have its own email address instead of borrowing a human's, what an address proves and does not prove, and how to make it trusted.

  • Human in the loop email for AI agents

    Keep a human in the loop for an AI agent's mail: held mail, drafts only, send approval and write levels, and why the control must sit outside the prompt.

Screening and email security

AI agent security

  • What is prompt injection?

    Prompt injection is text that makes an AI model follow an attacker's instructions. How direct and indirect injection work, and what reduces the risk.

  • Indirect prompt injection

    Indirect prompt injection hides instructions in content an AI agent reads. Why email is the most open channel, a worked example, and what reduces the risk.

  • The OWASP Top 10 for LLM Applications, mapped to email

    The ten risks in the OWASP Top 10 for LLM Applications 2026, each mapped to what it means when an AI agent has an inbox, with the 2025 numbers.

  • Zero-click AI attacks

    A zero-click AI attack needs no click because the assistant reads the message itself. EchoLeak (CVE-2025-32711) explained, and what changes for defenders.

  • AI agent security

    The main risks when AI agents act for you: untrusted input, excessive agency, data leaving and identity. A layered defense, and why the inbox matters most.

  • LLM data exfiltration

    How AI agents leak data: steered sends, links and images that carry data, replies to the wrong party. And the controls that limit what leaves.

Building with agent email