Over the last several weeks I've been my own personal AI financial advisor to help me trade and invest, budget, forecast finances especially now that I'm unemployed, and planning a move, and figure out international tax implications of all this. I've seen a lot of folks using Claude Code for finance related tasks, for summarizing, aggregating info and making projections. So I wanted to share my experiments and learnings.
It's grown into a pretty comprehensive set of skills and repos that know everything about my personal finance history. It started with me forking a few python projects that do fundamental and technical analysis (there were a couple big AI IPOs this year I was interested in). I added some rules, and memory to have it understand my investing thesis and risk tolerance so it would stop giving me generic advice and help with tax loss harvesting and prevent wash sales across exchanges. Then I kept going and built myself a copy of the budgeting apps like Monarch/Origin/RocketMoney to handle my sufficiently complex situation. I got one of my agents with computer use to extract data from legacy banking portals, label transactions, setup rules, and configure useful dashboards.
I had a lot of learnings along the way. Most of it came down to getting the right context, and using a better agent/model and computer use to get access to things not exposed via API.
Over the last several weeks I've been my own personal AI financial advisor to help me trade and invest, budget, forecast finances especially now that I'm unemployed, and planning a move, and figure out international tax implications of all this. I've seen a lot of folks using Claude Code for finance related tasks, for summarizing, aggregating info and making projections. So I wanted to share my experiments and learnings.
It's grown into a pretty comprehensive set of skills and repos that know everything about my personal finance history. It started with me forking a few python projects that do fundamental and technical analysis (there were a couple big AI IPOs this year I was interested in). I added some rules, and memory to have it understand my investing thesis and risk tolerance so it would stop giving me generic advice and help with tax loss harvesting and prevent wash sales across exchanges. Then I kept going and built myself a copy of the budgeting apps like Monarch/Origin/RocketMoney to handle my sufficiently complex situation. I got one of my agents with computer use to extract data from legacy banking portals, label transactions, setup rules, and configure useful dashboards.
I had a lot of learnings along the way. Most of it came down to getting the right context, and using a better agent/model and computer use to get access to things not exposed via API.