Aurelio is a public, MIT-licensed personal-finance application that runs on the user’s computer and keeps its records there. Its AI adviser is optional and connects through OpenRouter with the user’s API key. Developers can therefore clone and run the financial tracker locally while deciding separately whether to configure the external AI integration. The repository documents the license, local operation and optional adviser.
A ledger for assets, liabilities and goals
Aurelio records banks, brokers, cash and dated holdings. Its portfolio views calculate position costs, gains and dividends, while its fund look-through breaks exposure down by country, sector, company and currency. The app also covers recurring investments, homes, debts, income, expenses, savings rates and the returns required for financial goals. These functions are listed in the project README.
Running the demo requires Git, Node.js 22 or later and uv. The supplied macOS, Linux and Windows scripts start the app with an invented household and open it at http://localhost:8000. The repository provides the requirements and demo commands.
For non-demo use, records reside in backend/data.db. The project says it saves a copy beside that file before an update changes the data format. The storage path and pre-update copy behavior are documented in the README.
The repository also documents separate backend and frontend test commands using pytest and npm test, plus hot-reload commands based on Uvicorn and the frontend development server. Those development workflows appear in the project documentation.
The adviser introduces a second boundary
The “Ask Aurelio” chat answers from the user’s records, can search the web and presents proposed changes as cards that the user must confirm. Enabling it requires an OpenRouter account with credit and an API key stored in backend/.env; users can select a model in the chat. The repository describes the chat behavior and OpenRouter configuration.
Analysis: Local record storage and local AI processing are not the same property. Aurelio keeps its ledger on the user’s machine, but the optional adviser routes through OpenRouter and can use those records when answering questions. Developers evaluating the app should therefore treat storage and AI processing as separate trust boundaries rather than reading “runs on your own computer” as a claim that every enabled operation remains local. That interpretation follows from the project’s description of local records and its OpenRouter-backed adviser.
The confirmation-card flow provides a visible checkpoint before a proposed change is accepted. Analysis: It does not settle the separate question of what financial context is appropriate to submit to an external AI service. The unresolved trade-off is between the adviser’s record-aware answers and web search on one side, and keeping the application free of that external integration on the other. The underlying confirmation, record access, web search and optional-AI behaviors are documented by the project.
Top comments (0)