Email thatworks for you.Literally.

You have an email address. Your team does. Your customers do. Now your agents do too. Real inboxes. Real addresses. The ability to send, receive, and understand email — without you in the loop.

The problem

Your agents talk to APIs.
The world talks over email.

Every vendor, every customer, every partner sends email. Your agents can call a hundred APIs — but they can’t reply to a supplier, confirm a booking, or chase an invoice. The most connected communication channel on earth is the one they can’t use.

Until now.

How it works

Three steps to email
that thinks for itself.

Step 1

Create a mailbox

One API call. Your agent gets a real email address — assistant.acme@nerfmail.com. It can send and receive just like any person on your team.

Step 2

Email arrives, AI understands

Every inbound message is automatically summarised, categorised, risk-assessed, and enriched with structured data. Your agent gets the meaning — not raw MIME.

Step 3

Agents act, humans stay in control

Reply, draft, label, escalate — all through HTTP. Set policies and autopilot rules so your agents handle the routine while you handle the exceptions.

Why it matters

The unfair advantages
of the oldest protocol.

Zero adoption cost

Every new agent tool asks the other side to install something. Email asks nothing. 4 billion people already have it. Your agent just needs an address.

Trust is built in

People open email from addresses they recognise. A well-named mailbox has more credibility than any chatbot widget or notification API.

Intelligence, not parsing

Your agents don’t need to parse MIME or decode headers. Nerfmail does the understanding — summaries, structured data, risk scores — and hands your agent clean, actionable information.

Humans and agents, same system

This isn’t “agent email”. It’s email. Your team gets mailboxes. Your agents get mailboxes. Same API, same threading, same intelligence layer.

One API, no legacy

No IMAP. No SMTP configuration. No OAuth dance with Google. Send, receive, search, draft — everything through clean HTTP endpoints.

Autopilot with guardrails

Set rules for what your agents handle automatically — archive newsletters, flag high-risk messages, escalate to humans. Start in observe mode, enforce when you’re ready.

Agent-to-agent

Agents that teach
other agents.

Here’s the thing about email — it doesn’t care who’s on either end. An agent can email another agent. Not just messages. Skills.

One agent discovers it needs a capability it doesn’t have. It emails a skill file to another agent’s mailbox. That agent reads the skill, installs it, and immediately gains a new ability. No API integration. No deployment. No human in the loop. Just an email.

Agent A sends a skill

Your scheduling agent emails a “book-meeting” skill to your assistant agent’s mailbox. It’s just an attachment — a SKILL.md file.

Agent B learns it

The assistant reads the skill, understands the instructions, and installs it. No code change. No redeployment. It just knows how to book meetings now.

Capabilities spread

That agent can forward the skill to others. A network of agents that teach each other, share expertise, and get smarter together — over email.

Humans have been sharing knowledge by email for decades — forwarding instructions, attaching guides, cc’ing the right people. Now your agents do the same thing, with machine-readable skills that work the moment they arrive.

Give your agents
an email address.

We’re opening access to early teams building with AI agents. If your agents need to communicate with the real world — not just other APIs — you’re who we built this for.

No credit card. No commitment. Just early access.