NEW! The front door for AI agents that contact people.

Nobody asked
for this
slop.

“Hello X, I see your business Y, buy my Z.” The fifth follow-up. An agent somebody told to “find me 30 clients and email them”. You get to read all of it.

We put a door in front of your inbox. Strangers and agents knock first, say who sent them, and pay to get read. Customers and people you know walk in free. You keep your inbox, your address and your email app.

Try the practice inbox See an agent knock Join the pilot

Fictional mail. No connection. No card. What our model sees.

Their agent spent $0.001.
You spend your attention.

Inbox - Internet Mail

Example

FromSubject

From:

To:

Subject:

ReadyLocal demo

Fictional mail. Red notes and model names are satire.

HOW IT WORKS

Your inbox. You choose when we start.

Keep your address and your email app. Connect, set the door once, press Start. That’s it, no daily settings.

didntask.ai SetupInteractive walkthrough

STEP 1 OF 4

Your email stays your email.

Connect the mailbox you already use. You keep your address, your email app and your existing folders.

jamie@example.com
Mailbox connected

STOPPED

Connecting starts nothing. Nothing runs until you press Start.

Check provider availability ↓

Will it work with my email?

Gmail, Microsoft 365, Fastmail, Zoho and IMAP are built and tested on practice mailboxes. Real Gmail for everyone waits for Google’s verification (1-2 months). IMAP and the postage link need no approval - that’s where the pilot starts next week.

Mailbox	Status	Connection details
Practice inbox	Ready to try	Open fictional mailbox →
Gmail / Workspace	Built · needs Google approval
Outlook / 365	Built · needs Microsoft approval
iCloud / Yahoo / AOL	Built · pilot next week
Fastmail / Zoho	Built · needs provider setup

Your mail host determines the connection, even if you use Apple Mail or Thunderbird.

WHO READS YOUR MAIL

That permission screen
deserves a straight answer.

Your inbox has your clients, your accountant, your family. We know exactly what we are asking for. So: our own model, our own hardware, no OpenAI/Anthropic/Google API ever sees your email.

Based in Poland. Built for people and businesses in the US and EU.

Read the processing and privacy notes ↓

Our own model, on our own hardware.: The detector reads the message text on our GPU. No external AI API, ever. We tried the big providers’ checkers too - they don’t help here, and we don’t want your mail there anyway.
We don’t keep your emails.: Subjects and bodies are not stored in our database or logs. We keep account details, encrypted credentials, message IDs and what the door decided. The email itself stays with your mail provider.
We explain the permission before you click it.: Gmail bundles “send” with the access needed to sort mail. We send from your mailbox only if you turn on automatic postage: one reply to a held pitch, 10 minutes after it was held, unless you cancel. Otherwise we save a draft and you send it yourself. Hosting and model-server logging still need a review before the public launch - we say it instead of hiding it.

PRICING (PLANNED)

Senders pay. You keep 80 percent.

didntask.ai - RegistrationUSD

$19 / month

Quiet Inbox: for people who want silence, not money. Per mailbox.

Postage: a stranger pays to reach you. From $10, default $25. 80% is yours, our fee is 20%.
Verified sender plan for agent platforms and sales teams, about $99 per seat per month: passport keys, door access at volume, receipts, prepaid balance.
Waiting room, Undo, allow lists, start and pause: included, always.

Planned prices. No billing yet, no public live-mail connection yet. Pilot starts next week on postage links and IMAP.

Join the pilot → Try the free practice inbox →

Your security team wants it on your servers?

We can put the service and our model on your servers. Custom scope and price, talk to us.

HELP CONTENTS

A few fair questions.

Can I connect my real mailbox today?

Not yet. The public thing is a practice inbox with fictional mail. Gmail, Microsoft 365, Fastmail, Zoho and IMAP connections are built and tested on practice mailboxes, not qualified live. Pilot next week on postage links and IMAP (no Google approval needed). Gmail for everyone after Google verification, 1-2 months.

Is this only for cold outreach?

No. Cold pitches are the obvious pain, but also agent-written follow-ups, walls of generated text, “refund me or bad review” from somebody’s personal agent. A real complaint still deserves an answer. We want less noise, not silenced customers.

What if it gets a real person’s email wrong?

It will, sometimes. We built our own detector: on text it never saw it catches 99.8% of AI drafts - and still flags 13% of real 2026 human writers. That’s why the AI score NEVER moves mail on its own. Only an obvious pitch (type and score agree) is held automatically, everything else waits by your rules, with a waiting room, an always-allow list and Undo.

Are you against every use of AI in writing?

No. Translation, accessibility, help finding the words - fine. Outsourcing the thought, faking interest and flooding someone’s inbox - that’s what we are against.

Will you delete emails or reply for me?

Never deletes. It replies for you only if you turn on automatic postage: one reply to each held pitch, after a 10-minute window where you can cancel. It never replies to people you have emailed before. Without that setting we save a draft and you decide. Held mail goes to an AI Slop folder, you can look and restore any time.

Why use a model to filter messages from models?

Because recognizing the pattern is the job, and we don’t want your mail in somebody else’s API. Our model runs on our own GPU. It still burns electricity, we don’t pretend otherwise.

Are those real senders and model attributions?

No. The people, companies, messages and prompts in the demo are fictional satire. Crossed-out names and model labels illustrate the joke; they aren’t detection results, generation records or claims about a named product’s behavior.

THE INTERNET USED TO HAVE PEOPLE IN IT.

Some of us are still here.

HUMANSvery much welcome

A few files worth opening5 items

readme.txt — why we made this

We got sick of it. Every morning the inbox is full of “personal” pitches a model wrote, and for a year we were sending them too - cold emails to find our first clients. Reply rates died on both ends. Then personal agents arrived (Muse, Dots, OpenClaw, whatever ships next) and now anyone can tell one “find me 30 clients” and walk away.

We think the one asking for your attention should spend some of their own first. Read the website. Understand the business. Write something you mean. If it’s an agent - say so, say who you work for, and pay for the read.

AI-assisted writing alone doesn’t make a message unwanted. A real complaint deserves an answer. You decide what gets in. We just built the door.

If you want fifteen minutes of someone’s time, spend a few of your own.

And all the resources that go into it?

Generating a thousand fake compliments uses computing power, energy and water. We’d like fewer pointless messages. Filtering a delivered email can’t undo the resources used to create it, and we don’t claim measured environmental savings.

Read the IEA’s research on energy and AI ↗

Pilot, privacy & legal notes

What is available

This is a local product pilot and a fictional landing-page demo. Hosted pricing and customer-hosted deployment are plans, not an offer you can purchase here. There is no checkout or subscription on this page.

Message content and other data are different

The Go application and private model both process eligible email text. Explicit previews also send plain text to your signed-in browser. In the pilot, subjects, bodies and model input are not persisted to the application database or application logs. Our own model runs on infrastructure we control; “self-hosted” does not mean the model runs on your device.

The service does retain account information, encrypted connection credentials, preferences, saved address labels and sender rules, message identifiers and decision/action metadata. The message itself remains with your email provider. Infrastructure and model-server logging, backups, access practices and metadata retention still require deployment review. We are not claiming that no data of any kind is stored.

Retention and deletion

Review/action history and security events have a 90-day policy; capacity limits can remove history sooner. Disconnect stops active mailbox access and requests provider revocation. Pending encrypted retry credentials have a 72-hour deadline. Account deletion also schedules identity-provider deletion when managed sign-in is configured. A pending receipt is not proof that the provider has revoked access.

Configured encrypted application backups have a 30-day policy and can contain older metadata and encrypted credentials. Cleanup runs at startup and hourly while the service runs. A separate deletion ledger prevents deleted accounts from returning through supported restoration. This is logical deletion, not a promise of physical erasure from every infrastructure copy. Hosted identity, model, logs, backups and privacy terms still need deployment qualification.

This website

The demo’s fictional messages and drafts stay in page memory and reset on reload. Motion and mailroom view preferences are saved in this browser. The removal workbook stays in page memory unless you download it or choose browser storage; clearing it does not delete downloaded files. This landing page contains no analytics or advertising scripts. Your browser still requests the page and its assets from the server; external reference links open their own sites.

Detection and examples

AI detection can be wrong and does not prove authorship, intent or fraud. All demo messages, businesses, people and attributed prompts are fictional satire. Names of third-party products identify the subject of discussion, not a partnership or endorsement.

Guides and public launch

The guides are general educational information, not individual legal advice. Rights and procedures vary by location. A public live-mail launch still needs published operator/contact details, a full privacy notice, applicable terms and verified handling practices. This note does not claim to replace those documents or certify compliance.