What data would you give Claude/ChatGPT if you wanted truly deep stock research?

When you ask an AI agent to research a stock today, it goes off and searches the web — pulls a P&L number from one site, a balance sheet figure from another, maybe a stale ratio from a third. It works, sort of, for one company, one time. But the data is fragmented, often outdated, and the agent starts from zero on every question.

Now imagine the opposite: the AI has structured access to everything, all the time. Full profit & loss, balance sheets, cash flows, ratios, valuations, analyst targets — not for one company, but for the entire US market, 10,000+ companies, queryable in a single conversation. "Compare the margin trends of every mid-cap semiconductor company over 5 years" becomes one question, not an afternoon of searching.

That's what we built our MCP server for (shinobidata.com/mcp) — but honestly, we're more interested in your take: if your AI assistant had access to any financial dataset, what would you actually ask it?

And what data do you think is still missing for AI to do genuinely institutional-quality research?