Convert PST to MBOX — in your browser

MBOX is deceptively simple: one plain-text file holding every message in a folder, each one introduced by a line that begins with the five characters "From " (the From_ line). That simplicity is why nearly every mail client can import it — and it is also the format's one sharp edge.

Because a bare "From " at the start of a line marks a new message, any body line that genuinely starts with "From " has to be escaped or the mailbox splits in the wrong place. The variants exist entirely to disagree about this: mboxo escapes nothing reliably, mboxrd escapes such lines by prefixing ">", and mboxcl/mboxcl2 add a Content-Length header instead. Mailward writes the mboxrd style — the ">From " escape — because it is what Thunderbird and Apple Mail expect. Reading is the mirror image of that problem: Mailward only treats a line as a separator when it has the full From_ shape — a sender token followed by a seconds-precision date — so an mboxo file whose body says "From my point of view" opens intact, instead of splitting mid-message the way importers that key on the five-character prefix alone still do.

Converting PST to MBOX therefore means two jobs, not one: rebuilding each message as a MIME document from the PST's property sets, then framing those documents with correct From_ lines and escaping. Both happen locally, in a Web Worker, with nothing uploaded.

The separator line itself has a defined shape: mbox(5) specifies "From sender date" — three parts on one line, with the date in the asctime form classic Unix tools emit ("Thu Jan 1 00:00:00 1970"). Mailward writes exactly that, in UTC, and writes the sender as a whitespace-free token. That second point matters for Outlook archives specifically: mail that never left an Exchange server records its sender as an X.500 directory name ("/O=ORG/OU=…/CN=RECIPIENTS/CN=JANE DOE"), which can contain spaces — left alone, it would break the three-part shape that stricter readers, mutt among them, validate before accepting a file as a mailbox at all. Mailward collapses that whitespace in the separator; the real sender, untouched, stays in the message's own From header. The separator date is bookkeeping too — importers take the date you see from each message's Date header, and a message with no recoverable timestamp gets the Unix epoch rather than an invented one, so converting the same archive twice produces identical output.

Two details keep the splitting exact. Rebuilt bodies are quoted-printable encoded, and the encoder writes any body line that would start with "From " as "=46rom " — so framing and content are kept in different alphabets before the mboxrd escape is ever consulted. And the escape that remains follows the mboxrd rule precisely: a line matching any run of ">" followed by "From " gains one more ">", which is what makes the transform reversible on import. The simpler mboxo convention — escape only the bare case — cannot be undone without guessing, because a line that genuinely began with ">From " and an escaped "From " look identical afterward.

Output is produced one framed message at a time and encoded to bytes as it is generated. With attachments embedded as base64 inside each message, a folder's output can be several times its message text alone; incremental framing keeps peak memory near the size of one message rather than the whole folder.