Drawtree
Every investment thesis as a tree. Every claim has a kill condition. Every verdict is signed, timestamped, and disputable. The wire protocol for AI-native equity research.
How much should I buy?
Picking the stock is the easy part.
Type in the names you like, where you think they're going, and how sure you are. Out comes how much of each to buy, how much to leave in cash, and the orders to place at IBKR or Futu. Live prices load themselves.
How will you know you're wrong?
The day you bought, you had a reason. Is it still true?
Work with the AI you already use to break that reason into a tree — where every claim has to name what would prove it wrong. Can't name it, can't commit it. From then on it's re-checked weekly against fresh evidence, and you hear about it the week a verdict turns. Not when you're down 30%.
Works with Perplexity, Claude Desktop, any Remote-MCP client.
Already have an account? Sign in
How is this different from just asking an AI?
- Ask an AI about a stock and you get an essay you can't check. Here every claim must carry an observable kill condition — a number, a date, a disclosure. The validator rejects the tree if one is missing, so you cannot publish an unfalsifiable claim.
- The verdict is computed from the leaves, not asserted in a closing paragraph. You can see which specific claim moved it, and which one is worth watching hardest.
- A chat ends when you close the tab. A committed tree is re-checked weekly against fresh prices and new evidence, and tells you the week something breaks.
- Leaf questions are grounded in 164 strategy frameworks with canonical full-text — the AI picks the framework that fits your case and asks from the literature, instead of asking generically.
- Trees are content-addressed and append-only, so when you change your mind the old version is still there. Your past self stays on the record.
- 50 free credits on signup — enough to publish your first tree end-to-end. No credit-card prompt. Committed trees stay private to you.