The receipts are already in your inbox: the hosting bill, the flight confirmation, the subscription that renews quietly every month. What costs the hour is the download-rename-upload loop. TaxItEasy pulls them out two ways, both shipped: forward an email to a private address, or connect the mailbox read-only and let it check every fifteen minutes.
We sell the software this page describes, which is worth knowing before you reach the last section. Everything ahead of it is mechanics, including the parts where the answer is "that does not work, do this instead".
Can I automatically extract receipts from email?
Yes, and it needs no third-party connector. TaxItEasy runs two ingestion paths: a private forwarding address that accepts mail from any provider, and a read-only connection to Gmail, Outlook or generic IMAP. Both are included on every plan, Free included, and both end in the same Documents list.
"Automatically" has a boundary worth naming. The pipeline decides per email whether something invoice-shaped arrived, pulls the attachments out of the ones that qualify, and creates a document per attachment. It does not read receipts out of email bodies. A merchant who writes the amount into the message and attaches nothing produces a log entry and no document, so print that one to PDF yourself.
Quota is counted when documents are created, after classification, so a forwarded newsletter costs nothing and a mail with three invoices attached costs three.
Forwarding or a connected inbox: which should you use?
Forward when you can name the senders. Connect when you cannot. Forwarding is push based: the mail arrives, a webhook fires, the attachments enter the pipeline with no polling interval in between. A connected mailbox is checked every fifteen minutes and needs no rule maintained inside your mail client.
The honest trade-off is about access, not speed. Forwarding grants nothing ongoing: you decide per email, or one rule forwards exactly what it matches. A direct connection catches the vendor you forgot to put in that rule, the one that turns up missing in March, and the price is a read-only grant over a whole mailbox going forward.
Both can run at once: a per-message deduplication key means overlap never produces a second document. If you want an opinion, start with forwarding. It takes a minute, and you undo it by deleting a rule.
How do I set up the forwarding address?
Copy the address, point a rule at it, then forward one invoice as a test. In the app, open Settings, then Email integration, and copy the address shown at the top. It looks like [email protected], one address per company, and it works from any mail provider.
- Open Settings, then Email integration, and copy your
u-…@in.taxiteasy.orgaddress. - In Gmail, go to Settings, See all settings, Filters and Blocked Addresses, Create a new filter. Put a vendor in the From field, tick "Forward it to", add your address. Gmail sends a confirmation code there, so it turns up in your Activity log, not your inbox.
- In Outlook, go to Settings, Mail, Rules, Add new rule, then pair a sender condition with the "Forward to" action. Microsoft 365 tenants sometimes block external forwarding until an admin allows it.
- Forward one known invoice and open the Activity log. One new row, with sender, subject, classification and the number of attachments extracted, means the path works.
The hash in that address is sixteen hex characters derived from a one-way cryptographic hash, so it cannot be guessed from your company name or your signup date and cannot be reversed. Treat it like a password anyway: anyone holding it can push documents at your account. The second line of defence is sender authentication, since every inbound mail has to pass SPF or DKIM against the original sender's domain or it is rejected. More in set up email forwarding and the inbound address explained.
How do I connect Gmail, Outlook or IMAP instead?
One consent screen, no wizard. Open Settings, then Email integration, click Connect account and pick Gmail, Outlook or Generic IMAP. Gmail and Outlook route you through the provider's own consent screen. Generic IMAP asks for host, port, address and an app password. The first check runs on the next fifteen-minute cycle.
What the consent screen grants is read access to mail. Sending, replying, deleting, labelling and archiving are not in the scope, and the grant is printed in plain text on the screen before you click Allow. A server-side whitelist also rejects any token that comes back with wider permissions than we asked for.
Generic IMAP is the fallback for Fastmail, iCloud, your own server, or a Workspace tenant whose admin blocks third-party OAuth. Use an app-specific password. If your provider does not issue those, stay on forwarding rather than hand over your real credentials. The list is Gmail, Outlook and IMAP, and a mailbox that speaks none of them is out of reach. Steps in connect Gmail, Outlook, or IMAP.
What happens to emails that are not receipts?
They are logged and skipped. Every inbound email gets one of four labels, invoice, receipt, statement or irrelevant, from keyword checks on the subject, the body and the attachment names. Irrelevant means no document is created and no quota is used, and the decision is visible in the Activity log.
That is a fixed rule set with German and English keyword lists, not a model that learns your senders, and the predictability is the useful part. Any PDF attachment counts as at least a receipt candidate, which is why templated vendor mail lands correctly almost every time.
The weak spot is the minimal "see attached" mail: three words of subject, no keywords, a JPG photo of a till receipt. When one gets skipped, upload the file directly or forward it again with "invoice" in the subject. Nothing was stored the first time, so the retry is not blocked as a duplicate. What happens when a marketing PDF sneaks through the other way is in what happens when you forward a newsletter.
What does TaxItEasy read from your inbox?
Headers, one pass over the body, and the attachments. Sender, date and subject are stored for the Activity log. The body is read once for classification and then discarded, never stored permanently. Attachments go through the same checks as a manual upload before anything is kept.
Those checks are where most bad input dies. Accepted attachments are PDF, JPG, JPEG, PNG and HEIC, up to 20 MB each, and a magic-byte check compares the file's real bytes against its declared type, so a text file renamed to .pdf does not get through. Archives are not unpacked, on purpose.
What survives intake, and what to do with the rest, is in email attachment size and format limits.
On the connected path, we poll rather than mirror: no copy of your mailbox is created, and each check pulls only the attachments from the mail it classified as invoice-shaped. Disconnect is one click on the same page: the encrypted token is wiped in our database before any network call, then revoked at Google. Microsoft publishes no equivalent single-token revoke, so that path relies on the local wipe plus expiry. Servers sit in Frankfurt and the data at rest stays in the EU; the exceptions and the encryption details are on the security page, and the cleanup path is in how to disconnect an email account.
Where the extracted documents end up
An attachment that arrived by email is not a second-class document. It goes through the same intake as a file you drag into the browser, then through the extraction step that reads supplier, date, net, VAT and total off the page, and it lands in Documents with the original subject line as context. From there it can be matched against the bank line that paid it.
Which is the point of setting this up. Your inbox stops doubling as a filing system you have to remember to empty, and the quarter closes with the receipts already sorted instead of in a folder called Later. Photographed paper follows the same route, and whether you can bin the original afterwards depends on your own retention rules, which is worth settling before you start throwing anything away.
When your accountant asks for the period, the handover is access rather than a zip file. That side is walked in share receipts with your accountant. The rest of the capture story sits on the receipt scanning app page. TaxItEasy organises and prepares; your tax advisor reviews and files.
Frequently asked questions
Is extracting receipts from email included in the free plan?
Yes. Both paths run on every plan, Free included, and the gate that once restricted email sync to paid plans is gone. What a forwarded email costs you is document quota, counted per processable attachment, so one mail carrying two PDF invoices counts as two documents.
Does TaxItEasy read my whole inbox?
No. On the forwarding path it only ever sees the mail you forward. On the connected path the scope is read-only and the mailbox is polled, not mirrored: each check pulls attachments from invoice-shaped mail and leaves everything else alone. No copy of your mailbox exists on our side.
What file types and sizes arrive by email?
PDF, JPG, JPEG, PNG and HEIC, up to 20 MB per attachment, which are the same limits a direct upload gets. A magic-byte check compares the actual bytes against the declared type, so a renamed file is rejected. Archives, .msg wrappers and inline images never become documents.
Do you store the content of forwarded emails?
No. The body is processed once to classify the email and then discarded, so even a confidential forward leaves no body text behind. What stays is the sender address, the send date, the subject line for the Activity log, and the attachments themselves, which are encrypted at rest.
What if a real invoice is skipped as irrelevant?
Nothing is lost, it simply was not detected that time. Two recoveries: save the attachment and upload it directly, or forward the mail again with the word invoice in the subject, which the keyword check catches. The attachment was never stored, so duplicate detection will not block the retry.
Can I disconnect a mailbox later?
Yes, from the same settings page, and it takes one click. The stored token is wiped in our database immediately, then revoked at Google so it stops working on their side too. Documents already extracted stay where they are, and only future checks stop. Microsoft exposes no comparable endpoint.