Inside the app

AI assistant

A read-only "second founder" that knows your portfolio and never makes decisions.

The assistant is a chat dock inside the app that has read access to your real portfolio data — the scorecard, the playbook, the gate state, the economics of the product you're looking at.

What it's for

Open it when you're stuck. It's a coach, never a writer and never a decision-maker: it will point at the evidence you already collected, name the gate requirement you keep avoiding, and suggest which experiment to run next — but it can't edit anything, and it will tell you when data is missing instead of inventing an answer. Inside a product workspace it coaches that product; everywhere else, the portfolio.

The details

  • Streaming responses; the model is called through OpenRouter and every turn is traced (model, tokens, real cost) for observability.
  • Credits: one credit = one assistant reply. Each plan carries a monthly allowance, shown live in the dock footer. Errors after a failed call don't silently consume credits.
  • Pro/Team-gated — when your trial has ended you'll see an upgrade card (and a read-only banner) instead.

How to get value from it

Ask it the questions a good advisor would: "What's the weakest part of this scorecard?", "What evidence would move this product to BUILD?", "Which of my products is closest to its next gate?". It reads your actual data, so its answer is about your portfolio — not SaaS advice in general.

AI assistant dock open over a product report, answering an investor-pitch question with evidence from the workspace The dock coaching PageProbe: asked for a three-sentence investor pitch, it builds the answer only from evidence already in the workspace — and closes with the question an investor will actually ask. The footer shows remaining credits and the read-only promise.