Agentic Capital is an AI equity research system. We make the tool we wished we had when we sat on the buy-side: one that reads the filings, knows the history, and shows up at 7am with a brief instead of a blank page.
For thirty years, professional equity research has run on Bloomberg + Excel + a stack of broker PDFs. The data is good. The discipline is good. The price tag, and the user experience, are not.
Then large models showed up, and overnight everyone could ask the chatbot about NVDA. The answers were fast and confident. They were also, alarmingly often, wrong: the kind of wrong that costs money.
The bet behind Agentic Capital is that the bottleneck isn't the model. Vanilla Claude Opus 4.7 and our system run on the same base model. What's different is the harness: a direct data pipeline to FMP, an independent knowledge base of every name we cover, and pre-deployed research methodologies that tell the agent when to stop reasoning and start verifying.
A workstation that reads like a Bloomberg terminal and writes like a Financial Times analyst. Six core capabilities, one editorial register:
A small team. Buy-side and infra backgrounds. We dogfood the product every morning before market open: if the brief is bad, we ship a fix before lunch. The roadmap is set by our own research workflow, not by feature requests from a sales call.
Headquartered in San Francisco. Hosted on AWS, Vercel, and Supabase. Powered by Anthropic models via a multi-agent gateway. The full stack is auditable by enterprise customers on the Max tier.
We read for a living. Agentic Capital is what we read with.
Founding team · leon.zhao@agenticcapital.dev
We answer every email. Tell us what you're trying to do and we'll tell you whether Agentic Capital is the right fit.