Hey everyone! I’ve been a Tiller user for many years and my least favorite part has always been categorizing transactions. I generally leave transactions uncategorized for months and then I would have to spend hour categorizing them. AutoCat helps but doesn’t catch everything, especially new merchants or ambiguous ones.
As a software consultant in my day job, I built an AI agent that does it for me. It reads `(Blank)` transactions from my Tiller sheet, looks at my existing categories and past transaction
history, and categorizes new ones automatically. It writes the category back to the sheet and I don’t have to do anything.
The part I’m most proud of: it sends me a daily email digest listing what it categorized, and I can just reply to the email with corrections. Like if it put a Costco run under Groceries but it was actually for the office, I reply with 3: Office Supplies and it re-categorizes it. That’s the whole workflow.
Privacy note: I saw the conversation in the other GPT thread and I share those concerns. This runs entirely on your own machine. Your financial data never passes through my servers or anyone else’s. The only external call is to the AI model for categorization (you bring your own API key via OpenRouter, so you pick the model and control the relationship). I’m open to implementing an Ollama integration if people want to use a model local to their machine (or on a server you host). Everything else stays local.
It’s free for personal use: GitHub - jevy/bookkeeper-agent: Intelligent agent for bookkeeping when using Tillerhq · GitHub
It does require some technical setup (Docker, AWS for email, an OpenRouter API key), so it’s not a one-click install yet. If you’re comfortable with that kind of thing, I’d love for you to try it and tell me what you think. And if you’re not technical but this sounds useful, I’m working on a hosted
version that would be much simpler, there’s a signup link on the repo.
Happy to help anyone get set up right here in this thread.