# Ledger **Status:** SHIPPED, v0.7.5 on PyPI — https://pypi.org/project/arcaeon-ledger/ **Install:** `pip install arcaeon-ledger` **Non-proofs:** - truncation-without-witness - truth of content - authorship - default verify() still tolerates unchained rows before the first chained row (adoption-on-existing-log design) — pass strict=True to reject any unchained row, including a fabricated legacy prepend (0.5.4) - {'verify() returns a THREE-VALUED verdict, not a boolean': 'ok=true only when every row was checked; ok=false on a break; ok=null when the scan was bounded and cannot vouch for everything — an empty file (verified_scope "empty", since 0.5.8) or unchained rows skipped ("bounded_prechain_skipped", since 0.5.7). ok=null is falsy and is not a green; the CLI exits 3 for it. Code that branches on a bare truthy `ok` is reading a scope it was never shown'} - first_break names only the FIRST fault. `breaks` carries the total count, and a result that reports one named fault does not mean there was one fault Observability tools show you what your agent did. Ledger lets you prove it. Every action an AI agent takes is hash-chained to the one before it. Edit a row, delete one, or reorder history, and every later link breaks — verification names the exact line. You own a portable audit trail you can prove wasn't altered, and you can replay the exact state that produced a bad output. - Open-source core, free to self-host. Zero dependencies. - Hosted collection, long retention, and compliance export: coming. - Status: SHIPPED — v0.7.5 on PyPI. `pip install arcaeon-ledger` then `python -m arcaeon_ledger.selftest` to check the guarantees yourself. Contact: hello@arcaeon.io --- © 2026 Arcaeon™ · a sole proprietorship · terms: https://arcaeon.io/terms · privacy: https://arcaeon.io/privacy · practices: https://arcaeon.io/witness-practices · licensing: https://arcaeon.io/LICENSING · hello@arcaeon.io