Agrippa drafted De occulta philosophia in 1509-10, in his early twenties, and sent it to Trithemius for comment in April 1510. Trithemius praised it and told him to keep it quiet. He did, more or less, for two decades: the book circulated in manuscript and was not printed complete until 1533, at Cologne, in an edition that suppressed its printer and place because a Dominican inquisitor had denounced it while the type was still being set.
Its structure is the reason it mattered more than anything else in the field. Three books, ascending: natural magic, the virtues of stones, plants, animals and elements; celestial magic, the mathematical and astrological layer, numbers, seals and planetary hours; and ceremonial magic, the religious layer of divine names, angels and rites. Every correspondence table in Western magic since, Golden Dawn attributions, the planetary hours, the magic squares and planetary seals, the sigils of spirits, runs back to this book, usually through intermediaries who did not credit it.
The difficulty in reading him is that three years earlier he had published De incertitudine et vanitate scientiarum (Antwerp, 1530), a corrosive sceptical attack on the pretensions of every branch of learning, magic emphatically included, in which he disowns his own occult studies. Scholars have argued for four centuries about which Agrippa is the real one. The one fact that constrains the argument is chronological: he wrote the retraction first and shepherded the full magical encyclopedia into print afterwards, which is hard to reconcile with a straightforward conversion to scepticism. Christopher Lehrich's study treats the tension as deliberate rather than as a contradiction to be resolved.
He was a soldier, lawyer, physician, diplomat and perpetual debtor who quarrelled his way across Europe, and in 1519, as town advocate at Metz, he defended a woman accused of witchcraft and got her acquitted, attacking the inquisitor's reasoning and the Malleus it rested on. His student Johann Weyer carried that argument much further.