How it's built · no legalese

Your data lives on your computer.Here's the exact map.

By default, six things reach our servers — a subscription check, health counts, your delivered leads, short name-free snippets of what the AI skipped so its judgment can be checked, any teaching notes you write yourself, and the AI relay prompt itself (never stored). Switch to subscription mode and that sixth item disappears — the AI runs on your own Claude subscription instead, and post content never reaches us at all. This page draws the whole picture, claim by claim.

Get your first 7 leads free Read the formal policy

Where everything lives

Stays on your computer

Your Nextdoor login and browser profile
The full text of every post, if you switch to subscription mode — by default it passes through our AI relay to reach the AI provider, but we never store or log it, see below
Your local lead database
Your voice file and brand rules
Your Telegram connection token
Any edits you make and what you ultimately post

Reaches our servers

A subscription check (license key + device ID)
Health counts, scan timing, and error messages
For delivered leads: the public Nextdoor post link, the draft replies we wrote for you, and the full text of the post, the neighbor's own public words, which can include a name they wrote themselves
So you can review and correct the AI: for posts it read but didn't deliver, a short excerpt (capped at 200 characters) plus the AI's verdict, reason, and the post's public link, no name or neighborhood, auto-deleted after 90 days
When you teach the automation — from your dashboard or by replying on Telegram — the message YOU wrote is sent to us and stored to personalize your replies; neighbor post content is never included, and you can ask us to remove these anytime
By default, the AI relay prompt — the post text, and possibly a name or neighborhood the neighbor wrote themselves — passes through our servers on its way to the AI provider so it can classify the post and draft your replies; switch to subscription mode and this never reaches us at all

The third and fourth are where lead content reaches us, short excerpts only, no names, explained in full below. By default, the sixth call is our AI relay, which handles classification and reply-drafting — see the “phones home” section below for exactly what that sends and what we keep from it (nothing).

Most tools run in someone's cloud.Yours signs in from your house.

That single decision is why almost nothing needs to leave your machine, your login and your data stay where you are.

Signs in where you do

The automation drives a real browser on your own computer, on your own connection, the same place you'd sign in yourself.

Your login never leaves

You type your Nextdoor password into Nextdoor, on your machine. We never see it, store it, or ask for it.

Reads locally, keeps locally

Every post it scans lands in a database inside the automation folder. Delete the folder and that history is gone.

The AI, your call

By default, reading posts and drafting replies runs through NextJet's own AI relay — the post text reaches our servers only to be forwarded to the AI provider, and we never store or log it, covered in the 'phones home' section below. Prefer to run the AI on your own Claude subscription instead? Switch on subscription mode and that content skips our servers entirely.

Exactly six callsphone home.

Not “various diagnostics.” Six by default, each with a job. Switch to subscription mode and it drops to five — the AI relay call disappears entirely. Here's each one and exactly what rides along.

The subscription check

What it sends

Your license key, a device ID (your computer's name and network hardware address), an install ID, and the automation version.

Why it exists

It runs at every scan start to confirm your subscription is active, and the device ID lets us spot one key being shared across machines. No post content, ever.

Health check-ins

What it sends

Counts and timing, posts seen, leads found, scan duration, query performance, plus error messages when something breaks.

Why it exists

So we can spot a broken automation before you have to file a ticket. The automation strips content fields before sending, and our servers reject anything outside a strict, size-capped schema. Belt and suspenders.

Your delivered leads

What it sends

When a lead hits your phone, we mirror just enough to your dashboard to make it useful: the public Nextdoor post link, the draft replies we wrote, and a short excerpt of the post so you can see what the neighbor asked. Not their name or neighborhood.

Why it exists

So your dashboard shows real, clickable leads with the gist of each ask, while the neighbor's name, neighborhood, and full post stay on your computer and on Nextdoor. Full detail below.

AI-decision review snippets

What it sends

For the posts the automation read but didn't send you, a short excerpt (≤200 chars) plus the AI's verdict and reason and the post's public Nextdoor link, never the neighbor's name, neighborhood, or full post.

Why it exists

So your automation's accuracy can be checked and its judgment kept sharp (they're also included in your account data export). These auto-delete after 90 days and are never pooled into anyone else's automation.

Your own teaching notes

What it sends

When you teach the automation — from your dashboard or by replying to a lead alert on Telegram — the message YOU wrote. Never the neighbor's post, name, or neighborhood.

Why it exists

It's stored to personalize your replies, so the next drafts sound more like you. Used only for your own automation, and you can ask us to remove these anytime.

The AI relay (on by default)

What it sends

By default, every classification and reply-drafting prompt — which contains the post text and can include whatever name or neighborhood the neighbor wrote into their own post — is sent to our servers and relayed to an AI provider, Anthropic and/or OpenRouter (a model-routing service), using our own account.

Why it exists

It's how the AI reads posts and drafts your replies without you needing your own Claude subscription. We don't store or log what's inside the prompt, no database record of it, we only keep a count of calls and an estimated cost per license to meter usage. Prefer your own Claude subscription instead? Switch your automation to subscription mode and this call disappears — the prompt goes straight from your computer to Anthropic, never touching our servers.

Where lead contentis mirrored.

We want to be precise here, because a trust page that hides the exceptions isn't a trust page. The automation decides on your machine which posts are real leads. The winners, typically a handful a day, go to your phone on Telegram (or to Slack or email, if that's what you connected), and for those we mirror the public post link, the drafts we wrote, and a short excerpt of the post (capped at 200 characters) to your dashboard, so it shows real, clickable leads with the gist of each ask.

For the posts the automation read but didn't send you, we mirror the same kind of thing, a ≤200-character excerpt plus the AI's verdict and reason and the post's public link, so what it skipped and why can be checked and its judgment kept sharp. Those review snippets auto-delete after 90 days.

In both cases we deliberately do not store the neighbor's name or neighborhood, and we keep only the short excerpt, never the full post. The rest is someone else's personal information, and it stays on your computer and on Nextdoor (the link opens the real post). We don't sell, share, or advertise on what we keep; we use it only to power your dashboard and improve your own automation, never to train a model for anyone else. Delete your account and it all goes with it.

Built so it can't grow teeth

Excerpts only, never full posts
Delivered or kept-for-review, we store a ≤200-char excerpt, never the neighbor's name, neighborhood, or the full post.
Review snippets are capped + auto-deleted
The skipped-post snippets are capped per scan, used only to improve your own automation, and auto-delete after 90 days.
Rate limited per license
Every endpoint throttles each key, sized for normal scanning, not a firehose.
Deleted with your account
Ask us to delete your account and everything mirrored goes with it.

What we never have.Cancel-proof, by design.

Your Nextdoor username or password

The neighbor's name or neighborhood as separate saved data, we don't pull those out (a delivered post may still quote a name the neighbor wrote into it)

The full text of posts the AI read but did NOT deliver, for those we keep only a short excerpt (capped at 200 characters) plus the AI verdict

The stored contents of an AI relay prompt — it passes through momentarily on its way to the AI provider, but we don't log or save what was inside it

Your final sent text, your edits, or whether you sent anything (unless you choose to paste a reply into the teach box to train your voice)

Your browser profile or local lead database

The voice file your replies are trained on

Your Telegram connection token

Your full card number, billing runs through Stripe

And if you cancel?

The automation stops scanning, but the folder on your computer, the lead history in it, and every reply you ever sent stay yours to keep or delete. Email jayden@thenewworldagency.com and we'll delete your account data, mirrored leads included.

Built like it's your business.Because it is.

Questions about any of this? Email Jayden directly, or read the formal version, it says the same thing in longer sentences.

Get your first 7 leads free Privacy policy

No card required. Cancel in a few simple clicks.