Product
Curtis Audit.
Receipts for agent activity. Every AI inference that touches your notes gets an append-only record — sequenced per agent, checkable at any time, kept as markdown files you own.
Audit answers the question the other tools dodge: what did the agent actually do? When Curtis reads a note, rewrites a paragraph, or files an answer, Audit writes a receipt — what ran, what it touched, when, with which model.
How it works
- Append-only — receipts are never edited or deleted in place; the record only grows.
- Sequenced per agent — each agent's activity is ordered and checkable on its own.
- Markdown files you own — receipts shard into per-day files under
Audit/ledger/, and a rolling digest summarizes the ledger — plain markdown, readable in any editor.
Deliberately non-cryptographic
The ledger makes tampering visible — sequence gaps and appended-after-the-fact entries show — without pretending to be a blockchain. It's a paper-ledger ethic: honest records, not hash tournaments.
Shell receipts
Outside the plugin, a one-line shell wrapper files receipts into the same append-only ledger — so your own scripted agent activity lands in the same place. See the Audit page for the mechanism.