A four-month personal experiment turned into an open-source project this week. Ray is a terminal-based AI financial advisor that connects to your bank accounts, stores transaction data on your own machine, and strips identifying information before anything reaches an AI model.
The tool connects via Plaid, the bank-connection service used by most US fintech apps, and keeps everything in an encrypted SQLite database on your local drive. Before sending a query to an LLM, Ray redacts your name, merchant names, and transaction specifics - the model sees something like "transaction at [MERCHANT] for $42" rather than anything that could identify you or trace your spending.
The privacy-first design is the main differentiator from commercial alternatives. Most AI-powered financial tools - Copilot, Monarch Money's AI features, others - send raw transaction data to cloud servers. Ray keeps everything local, with personal information stripped at the query level before the AI ever sees it.
The obvious limitation: this runs in a terminal, which requires command-line comfort. That rules out most people who would benefit most from conversational money management. And connecting Plaid credentials to any tool - commercial or open-source - deserves real scrutiny. The advantage here is that the source code is public, so you can read exactly what the tool does before linking your accounts.
For developers and privacy-focused users who've been waiting for a financial AI that doesn't require trusting a startup with bank access, Ray is worth evaluating. The creator is actively soliciting feedback, which is a reasonable sign it's still early and rough around the edges.