What we do with your data

The short version

Fried Mango works your job search out of a mailbox made for it. It holds what you tell it, what it finds in that mailbox, and what it writes on your behalf. It does not sell any of that, does not advertise against it, and does not put it into a model that serves anybody but you.

What you sign in with

Google sign-in. We receive your email address and a Google-issued identifier for your account. We never see your Google password, and we do not store one.

What you give Google permission for

Nothing beyond who you are. We ask Google for your email address and an identifier, and for no access to any Google mailbox. There is no permission screen because there is nothing to permit.

Until 2026-10-10 this section listed two Gmail permissions and a count of your messages taken at sign-in. Both are gone: the mailbox your Search Team works is not a Google one.

Your job-search mailbox

We make a mailbox for you when you sign up, under your own name on our mail service, Sand Turn. It is the address on your resume and your applications, the address employers reply to, and the only mailbox your Search Team reads from or sends from. It holds your job search and nothing else, because nothing else has ever been sent to it.

Reading your mail is something the product can do, and here is exactly what that means today. One task reads — it looks at messages from the last thirty days and writes a line saying who each one was from and what its subject was, so you have a record. A job board's alert has its job links taken. A message from an employer you have applied to is opened and shown to the model, which says whether it is an interview request, a "we'll be in touch", or a no — and that one word moves the application along on your pipeline. A message it cannot place against one of your applications is written down and left for you. It does not reply, and nothing in a message is ever treated as an instruction.

Mail can go out three ways, and all of them are yours. One task sends applications: it attaches the resume you opened and the letter you read, to an address for an employer you said yes to. One task sends you a report: each morning, from your own address to your own address, a summary of your own pipeline that goes to nobody else. One task sends you one line the moment something needs you at once — an employer asking to interview you, or an employer's application form waiting for one tap from you — from your own address to your own address, and a copy to any other address you have added on your profile page and verified with a code. None can send anything else, and nothing goes to an employer you have not approved, yourself or by a standing yes you set in advance. Each was asked for first — authorization versions 3, 5 and 8 — which is the promise that matters.

Nothing reads your mail unless you have connected the mailbox from your dashboard, and you can ask us to disconnect it at any time.

We use your mail to run your job search and for nothing else — in particular, it is never used to build or improve anything that serves another person.

What we store

Things you tell us. Your display name, pronouns, timezone, the area you are looking in, and how far you are willing to travel. Answers you give about your work history and what you are looking for.

Your prospects — the jobs you are pursuing. Postings, employers, where each one stands, your notes, and a record of what changed and when. We say prospect rather than application because most never become one: a prospect can be a lead you are still weighing up, or a conversation with somebody you already know rather than a form to fill in.

Things written for you. Resumes, cover letters, application answers and the text of emails — each version kept, so it is always possible to say exactly what you approved.

Contact details for people at employers. Names, roles, email addresses and phone numbers of recruiters and hiring managers connected to your prospects. These are people who have not agreed to anything with us; we hold what is necessary to correspond with them about you, and nothing more.

Bookkeeping. When you signed in, whether you were admitted or turned away and why, which version of which document you accepted and when, and — for anyone who signed up before 2026-10-10 — the size of the Gmail mailbox that was measured at the door then.

Who else can see it

Fried Mango runs on infrastructure other companies operate, and each of them can technically reach the part it holds. They are all in the United States, and they are:

The first three are suppliers who hold your data because they run the machinery. None of them may use it for their own purposes. Ask us who they are and we will tell you — this is a description of roles rather than a secret, and the list changes as infrastructure does.

The browser hosting provider sees rather than stores, and it is new in this version. When an application goes through an employer's own form, the browser that fills it in runs on that provider's machines rather than ours, and reaches the employer's site through an address the provider supplies, because employers' sites turn away a browser coming from a cloud address before it has typed a thing. So it sees the pages that browser visits and every value typed into them — your name, your contact details, your answers, the resume attached, and the password typed into an employer's sign-in where an account was made for you. Nothing else is sent to it: not your mailbox, not your profile, not the key that seals that password. Each application gets a browser session of its own, started for it and ended when it ends, so nothing carries from one application to the next. The provider keeps a recording of each session for thirty days, which is how we can see what went wrong on a form, and its published policy says it does not use what passes through its browsers to train generative models without its customer's consent, which we have not given. It is described by role like the other three, and we will name it if you ask.

Anthropic is named rather than described, because it is the one that reads rather than stores, and the next section is about that.

The model, and what reaches it

Your Search Team is a set of language models. Those models are made and run by Anthropic, a company in the United States, and reaching one means sending it text.

What goes to them:

• Your profile answers — your work history, what you are looking for, what you will not take.
• The job postings being considered for you, and what we know about the employers.
• The messages in your job-search mailbox, where a task needs to read one to do its job.
• Drafts written for you, while they are being written.
• What you write to your Coach, with what it needs to answer you. The Coach is the part of your Search Team you talk to; the conversation is yours and goes nowhere else.

What never goes to them: the key to your mailbox, your passwords, or anything else that would let somebody holding it sign in as you. That is not a policy, it is how the system is built — the part of Fried Mango that talks to a model is refused those credentials outright, and there is a test that fails if that ever stops being true.

What Anthropic may do with it:

• They do not train models on it. Their commercial terms for the interface we use say so directly, and that is the basis on which we send anything at all.
• They delete it within thirty days. That is their standard retention. It is not zero — we have no arrangement with them for zero — so for up to a month a copy exists on their systems as well as ours.
• One exception, which we would rather you heard from us. If something we send trips their abuse detection, they may keep it for as long as two years while they look into it. We do not expect that to happen to a job search. We are telling you because a promise of thirty days with a footnote is worse than a plain description of both.

They are the only one. No other model company receives anything, and adding a second would be a new version of the [authorization](authorization.md) and a fresh acceptance — not a notice, and not a quiet edit to this page. That is why they are named: a name is something you can look up, and "an AI provider" is not.

What goes to employers, which is the point

The companies above hold your data because they host us, or read it because they run the model. Employers are different: they receive it, because putting your application in front of them is what you are here for.

When Fried Mango applies for a job or writes to a recruiter on your behalf, what goes out is what you would send yourself — your name, your contact details, your resume, your answers, and the text of the message. It goes from your mailbox, as you, to the employer you chose.

Nothing goes out until you have read the exact words and said yes. Not a summary, not "an email like this one" — the actual text that will be sent. If you set a standing yes, the words are still the exact words on file, written for you to read; what you have done is say yes to them in advance, up to a number a day, and each one is marked so you can withdraw it. That is the promise in the [authorization](authorization.md), and it is the one that governs everything in this section.

Once something is sent, it is theirs. An employer holds your application under their own practices, for as long as they keep it. We cannot recall it, delete it from their systems, or tell you what they do with it. That is true of any job application; it is worth saying plainly because here something else did the typing.

Applications are sent one at a time, each one approved by you — by email where there is somebody to write to, and through the employer's own application form where that is the only door. What goes is the exact resume and the exact letter you read; the file is made once and stored, so what an employer opens is the document you opened rather than a fresh version of it. On a form, every answer comes from what you have told us, and a question your profile does not answer stops the application and comes to you. Where an employer's system will not take an application without an account, one is made for you under your own job-search address, with a password made by code and kept sealed under a key held outside the database. What is stored is the employer system's hostname, your address, the sealed password, and when it was made and last used. The model that works the form is never shown the password, no person here can read it, and the only process that can open it is the one that types it into that employer's sign-in. You are told on the prospect's page when an account was made and where. Version 6 of the authorization is where you agreed to this.

When a page needs a person, you are shown the page. Some employers' sites put a puzzle in the way that is meant to tell a person from a program, and the browser cannot get past every one. When that happens in the middle of an application, the live form — the browser as it stands at that moment, with your application in it — is shown to you on that prospect's page in your own account, which nobody but you, signed in, can open; you tap through the puzzle there, and the application carries on from where it was. The address of that live view is kept in our database and sent nowhere: the line that tells you a tap is needed points at the prospect's page, not at the browser. The view is open for ten minutes. If nobody comes, the browser is closed with nothing submitted, the application waits for another try, and that try asks you again.

What we do not do

• No person here reads your mailbox — not for support, not for debugging, not out of curiosity.
• Nothing you give us trains a model that serves anybody else. What we hold makes your own search better and nothing wider than that: it does not go into a general model and it does not improve another candidate's Search Team. Anthropic does not train on it at all.
• Nothing is sold, rented, or shared with advertisers. There are no advertisers. You are the customer, or you will be; that is the whole point of charging you rather than an employer.
• We do not touch any mailbox but the one made for you. Not the account you sign in with, and not any other.

Keeping it separate is the protection

Your job-search mailbox is one we made for you, and that is the largest privacy measure here. A mailbox made for this contains this. Your real account contains your life, and we cannot read it: we hold no permission to, and we do not ask for one.

That got more important with version 2, not less: what we can read is also what we can send to a model, and the smallest mailbox is the smallest version of both.

How long we keep it

While your account is open. If you ask us to delete it, everything belonging to you goes, and it takes what is under it with it — your prospects, your documents, your notes.

Two things survive that, deliberately: the record that an email address signed in and was turned away, so a refusal does not have to be re-derived, and the record of which terms you accepted, which is the evidence that we asked. Ask for those to go as well and they will.

Deleting your account here does not reach back into what was already sent. Anthropic's thirty days run from when we sent it, not from when you ask us; an employer keeps what you applied with under their own practices. Both are described above so that "everything goes" is not read as more than it is.

Ask us to disconnect your mailbox and everything stops; it is done the same day. Your sign-in with Google grants us nothing a Google permissions page would list, so there is nothing to revoke there.

Getting a copy

Ask, and we will give you a copy of what is yours in a form you can read.

If this changes

We will publish a new version. Where a change is about something new being done on your behalf, you get asked again rather than told — that promise lives in the [authorization](authorization.md) and it is the one that matters.

Asking us something

support@friedmango.com. A person reads it.