Klavissimo doesn't import a PDF directly (a deliberate limit, see 4.1). For digitally typeset PDFs — publisher or course PDFs, not phone photos — there's a detour back to real notes: optical music recognition (OMR) via Audiveris, with MuseScore for the correction pass. Not a one-click route, but reliable if you take the time to check it.
Let Audiveris recognise the PDF
Audiveris (open source) reads the PDF and produces a file with its best guess at the notes — similar to text recognition for scanned book pages, just for sheet music.
Check errors against the original
Recognition isn't perfect: in dense passages with lots of beamed notes, Audiveris often mixes up note values, which makes bars look over- or underfilled. Open the recognised file in MuseScore — it flags the faulty bars itself — and correct them against the original PDF.
Export it cleanly
After correcting, export it just like a piece you set yourself, via File → Export → MusicXML — the steps for that are in the MuseScore guide.
Good to know:
To be honest, this route takes patience: triplets and ornaments are the most common source of errors, and hand scans or crooked photos are recognised noticeably worse by Audiveris than cleanly printed PDFs. For a well-known piece, it's often worth checking public-domain sources first before recognising it yourself.
Your data:
This route stays local too: Audiveris and MuseScore run as programs on your own computer — nothing from your PDF leaves it for the later import into Klavissimo.