What changed, what broke, and what we learned closing our own books. Written when there's something to say rather than on a schedule.
RSSThe Clerk.AI companion has been through App Store review and is now available to download. It is free for everybody. Clerk.AI usually has its first question for you within a few minutes of starting, which is far too long to sit and watch for — so you set a close off on the Mac, go about your day, and answer on your phone as the questions arrive. Some send you back to the Mac to sign in somewhere; many more are a correction or a clarification you speak into the companion to unblock it.
Read it →The biggest Clerk.AI release since the close itself. An iPhone companion brings the questions a close needs answered to your pocket — answer by speaking, with the clip transcribed on your own Mac. First run was rebuilt around a single question about your business, and what you watch while a close runs is the letter to your accountant filling itself in. Photos that look like receipts are offered to you to include or drop, and a period stays open after you send it, so when your accountant asks for one more thing it fetches that and remembers. Pricing became pay-what-you-decide, and the Mac app can now sign in and read what your licence grants.
The monthly close — previously a bundled Claude Code skill you ran from the terminal — now has a front-and-center UX inside the app: pick a month, quarter, or year, and Clerk.AI collects everything your accountant needs for that period's books, asks you only the questions it can't answer itself, and hands you a ready-to-send packet. Plus calmer defaults for a bookkeeping vault and a dramatically lighter background footprint.
Clerk.AI was born in a 2016 startup accelerator program. The first commercial release in 2017 — a hosted web app for Swedish freelancers — hit the last-mile audit-quality wall and got shut down. v0.1 (2019/2020) pivoted to local Python tooling, used by a small circle. v0.2 is the original 2016 vision actually shipping, with browser automation and local LLMs that finally work — going public again.