release1

← All products

POC

UI-Less Idea Evaluator

Evaluate a product idea in one guided AI session: a Lean Canvas, an 8-criterion scorecard and a Go, Refine or No-go verdict. Free, in Claude, ChatGPT and Claude Code.

  • ideas
  • ai
  • mcp
  • claude

The idea

Most ideas are never tested, because testing one properly means a canvas, a business model and a hard look at the weak spots. The evaluator runs that as a conversation. It interviews you, drafts each stage for you to correct, and pushes back on vague answers and flattering scores.

It asks one question of every idea: could this ship as an AI-native product, one that runs inside the AI its customers already use, with no app to build? And how would it make money when every file can be copied?

How it works

You add one connector to the AI app you already use. It gives that AI three tools, and the AI runs the conversation with you:

  • The rubric. The interview script, the Lean Canvas guide and the scoring anchors, so every AI asks the same questions and scores against the same scale.
  • The check. A server validates your finished evaluation and computes the verdict from fixed rules. The AI can't pick a verdict that disagrees with them.
  • The report. Your Lean Canvas, scorecard and a shareable one-page report.html, rendered and handed back.

One session, five stages

  1. Set up. Name the idea in a line.
  2. Lean Canvas. Customer segments first, then the rest, with a challenge question when an answer is thin.
  3. Business model. What you charge for, to whom, anchored to the alternative.
  4. Scorecard. Eight criteria scored 1 to 5, each with evidence from what you said and an honest confidence level.
  5. Verdict. The weakest criteria, what would lift them, and one concrete next step.

The scorecard

Maximum 40. Knowledge-based, skill-defined, verifiable outcomes, context-portability, low UI-dependency, knowledge stability, distribution reach and gated value.

The verdict is computed, not opinion. A score below 3 on verifiable outcomes or gated value caps the result at Refine, however high the total is.

Connect it

The connector address is the same everywhere. It needs no sign-up and no key.

https://uiless.release1.dev/mcp

Claude (web, desktop and mobile)

  1. Open Settings, then Connectors.
  2. Choose Add custom connector.
  3. Name it UI-Less Idea Evaluator, paste the address above and add it. Leave any sign-in fields empty.

Claude Code

claude mcp add --transport http ui-less-evaluator https://uiless.release1.dev/mcp

ChatGPT

Custom connectors in ChatGPT need developer mode, which depends on your plan. In Settings, open Connectors, then Advanced, turn on Developer mode, create a connector with the address above and choose no authentication.

Then ask for it

Start a new chat and say:

Use the UI-Less idea evaluator to evaluate my idea: a tool that ...

The AI will fetch the rubric and take you through the stages. At the end you get the three files back. In an app that can't save files, ask it to show report.html as an artifact or download.

What you get

Three files: report.html (a one-page visual you can share), lean-canvas.md and scorecard.md.

After a Go: the pitch pack

An evaluation tells you whether to build. The next step, with the same connector, turns a Go into material you can put in front of people:

  • Market research with sources: likely clients, market-size inputs, competitors and why now.
  • A market model that sizes the market and projects annual recurring revenue from your own pricing. The server does the arithmetic and flags any input you haven't marked as sourced or assumed.
  • An investor deck, a client deck and a user-journey deck covering your key personas.

Two things to know about how it works:

  • The research is done by your AI app, using its own web search, and it cites what it finds. The connector gives it the method and checks the result. Always read the sources before you rely on a number.
  • Decks are made as slides when your AI app can create them, and checked against the slide format. If it can't, you get the same decks as Markdown files instead, with one slide per section.

When you have a Go, ask:

Now pitch this idea with the UI-Less tools.

Figures are estimates. Anything the AI couldn't source is marked as an assumption, and details such as your team, the raise and the use of funds are left as bracketed placeholders for you to fill in. The connector lists every one still open.

Questions

Is it free? The evaluator is. Your AI app has its own plan or usage costs.

Does my idea leave my computer? Your conversation goes to your AI provider, as it always does. The connector only ever receives the finished evaluation, model inputs or deck to check and format. It stores nothing and logs no content.

Will it just say my idea is great? No. Every score needs specific evidence from what you said, and a score backed only by "I think so" is marked low confidence.

What kind of ideas is it for? The scorecard is tuned to ideas that can ship as an AI-native product, with no UI of their own. It tells you plainly when an idea is a better fit for a conventional app.