# What connecting your mailbox actually allows

> Most tools claim narrow access to your email. That is rarely how the permission works. Reveal's is wide, what we do with it is narrow, and you see both.

By [Miles Udemezue](https://www.linkedin.com/in/ude), Reveal. Published 2 November 2026, updated 4 November 2026.
Canonical: https://www.revealos.com/insights/what-connecting-your-mailbox-allows

When a client connects their mailbox to accounting software, the permission that gets granted is almost always wider than the thing the software says it does. Reveal is no exception, and rather than describe our access as narrow when it isn't, here is the honest version: the permission covers the whole mailbox, what we search for is narrow, and every search we run is written down where the client can read it.

This page is written so an accountant can send it to a client, or a client to their IT team, without anyone having to take a vendor's word for the shape of it.

## The permission is wide. Here is how wide.

It is the same shape at every provider we connect to. The exact permission differs; the breadth does not.

| Provider | What Reveal asks for | What it covers |
|---|---|---|
| Google | `gmail.readonly` | The entire mailbox. Not an inbox, not a folder, not a filtered view. Spam, the bin and mail the client has sent are all inside it. |
| Microsoft | `Mail.Read`, plus `offline_access` | Mailbox-wide read. `offline_access` is what lets the connection refresh itself rather than asking the client to sign in again every hour. |
| IMAP | No permission at all | A stored password, which is a different thing entirely. See below. |

None of these providers offers a per-sender or per-folder read scope. That is worth sitting with, because it means **no vendor can honestly claim their access is limited to particular suppliers at the permission level.** Any that do are describing their own software's behaviour, not the permission the client actually granted. The two are only as connected as the vendor chooses to make them, and you cannot see that from the consent screen.

IMAP deserves separate mention, because it is the one people assume is the safer option and it is the opposite. There is no scope and no consent screen: the client hands over a password, and anything holding that password can do whatever the account can do. The restraint is entirely the software's, not the protocol's. It also does not revoke the way a permission does. Disconnecting an IMAP mailbox in Reveal stops us using it, but the password still works at the email provider, so a client who wants it truly dead should change it there.

So the first thing to be clear about is the thing most vendors blur. **Access is wide for everyone in this category.** The questions worth asking are what the software does inside it, what it keeps, and whether you can check.

## The permission is wide. What we do with it is deliberately not.

This is the part worth being precise about, because "we have access" and "we use access" get collapsed together constantly, and the difference is most of the actual risk.

Reveal does not index your mailbox. It does not copy it, scan it in the background, or build a picture of it to query later. There is no standing process reading your email.

What happens instead is narrow and reactive. Your accountant is chasing a specific transaction: a payment to a named supplier, on a known date, for a known amount. That, and only that, is what gets searched for. The search is tied to the transaction. No transaction, no search.

Then the restraint continues past the search:

- **The documents that matched the transaction are kept.** That is the whole point of the exercise.
- **Everything else the search touched is not.** We hold the sender, subject and date, and nothing of the body.
- **Anything opened and not sent to your accountant is deleted after 30 days.**

So the honest shape of it is: the permission is mailbox-wide because no provider offers anything narrower, and the use is transaction-shaped because that is all the job needs.

**Today that restraint is ours.** It is a property of how Reveal is built, and you are taking our word for it, checked against the record of what we actually searched.

**In the next version it becomes yours.** Nothing in the mailbox will be searched until the client has chosen where Reveal may look: any sender, or only the suppliers they name. That is the difference between a vendor promising to be careful and a client setting a boundary the vendor cannot cross. We would rather be judged on the second, which is why we are building it, and we are not going to describe it as finished before it is.

## What read-only actually rules out

Every one of those scopes is read. None of them can send, change, label, archive or delete anything. We never asked for a permission that could, at any provider.

That matters more than it sounds, because the failures people imagine when they think about email access are mostly write failures: something sent in their name, something deleted, a rule quietly added. None of those are available to a read-only permission, whatever the software intends.

## What we actually search for

When your accountant is chasing evidence for a transaction, Reveal searches the mailbox by supplier name and date, and reads what comes back to check whether it is the right document.

That search runs at the provider, against the whole mailbox, because that is the only place the search can run. What comes back is a candidate list. A person on the accountant's side then looks at the candidates and decides.

## What we keep, and what we do not

The documents that matched a transaction are kept, because that is the point: the invoice goes to the ledger with the payment it belongs to.

For everything else the search touched, only the sender, subject and date are retained. The body is not kept.

Anything we opened and did not send to the accountant is deleted after 30 days.

## Protected subjects are flagged, not blocked

This is the part most worth reading carefully, because it is where our wording differs from what you might expect.

A client can mark subjects as protected: health, legal matters, anything personal. Reveal flags messages matching those subjects. **It does not block them.** A flagged message is still inside the mailbox the permission covers, and the system records whether it was opened.

What the client gets is not a guarantee that nothing was looked at. It is a record, every month, of whether anything was, and which. That is a weaker promise than "we never searched it" and a stronger one than a guarantee nobody can inspect, because a promise you can audit is worth more than one you have to believe.

If a tool tells you it blocks access to certain subjects at the mailbox level, ask how. With a mailbox-wide permission and server-side search, the honest answer is usually that it filters results after the fact, which is what flagging is.

## The record is the actual control

Every search Reveal ran and every document a person opened is written down, and the client sees it monthly on their own page. Not a summary that the work was done properly. The searches themselves.

That is the thing we would point an IT reviewer at first, because it is the only one of these claims that does not require trusting us. A permission is a promise about capability. A record is evidence of use.

## Credentials and disconnecting

Where the connection is a permission rather than a password, the client signs in at Google or Microsoft. Their password never reaches us, and we never see it.

Disconnecting in Reveal ends our access at Google. Microsoft also keeps its own record of the permission, and it stays listed there until the client removes it in their own account, so a client who wants it gone in both places should do both. For IMAP, as above, disconnecting stops us using the password but does not change it at the provider.

Where a supplier portal needs a stored login rather than a permission, those credentials are encrypted, used only at the moment of signing in, and never displayed to anyone. Reveal signs in and downloads invoices. It does not change anything in the account.

## What to ask any vendor, including us

Five questions that separate a real answer from a reassuring one:

1. **What scope do you request, by its exact name, at each provider?** "Read-only access to email" is marketing. `gmail.readonly` or `Mail.Read` is a fact you can check against the provider's own documentation. Ask about IMAP separately: a stored password is not a scope.
2. **Can the permission write?** If it can send, delete or modify, everything else is a matter of trust in their code.
3. **Where does the search run, and over what?** If it runs at the provider across the mailbox, say so. A vendor claiming per-supplier access at the permission level is describing software behaviour, not permissions.
4. **What is kept, and for how long?** Specifically: what happens to the things you looked at and did not use.
5. **Can the client see what you did?** Not a policy. The actual list of searches and opened documents.

We would rather be judged on those than on an assurance that our access is narrow, because it isn't, and neither is anyone else's.

## Why we publish it this way

Accountants are right to be careful here, and the clients who hesitate are usually not being difficult. They are asking a reasonable question that the industry has got into the habit of answering with a softer version of the truth.

A client who is told access is narrow, and later finds out the permission covers their whole mailbox, is entitled to assume the rest was spun too. We would rather hand over the wide version at the start, with the record that lets them check it, than win the conversation and lose it later.

If you want the underlying detail, the [Trust Centre](https://www.revealos.com/trust) has the per-source breakdown, and [the client page](https://www.revealos.com/clients) is written for the person being asked to connect rather than the person asking.

## Sources

- Reveal's implementation, as built, November 2026: requested scopes, retention, and what the monthly record contains.
- [Reveal Trust Centre](https://www.revealos.com/trust): per-source access, AI providers, retention.
- [For clients](https://www.revealos.com/clients): the same ground, written for the client.
