The Idea Evaluator
Connect

Is your project idea worth building?

Add the Idea Evaluator to the AI you already use, and it marks your idea like an exam answer instead of cheering it on: a score out of ten, the real risks noted in the margin, live research on who is already doing it, and, if it holds up, a plan and the documents to start building.

Connect your AI

Works in Claude, ChatGPT and Gemini. No account needed.

Use it in your own AI

Add the Idea Evaluator to ChatGPT, Claude, or Gemini as a connector. Your AI marks the idea with the same method: the six criteria, live research on what people are saying, a plan sized to your time, and the documents. Every stage is saved here, with a link you can share.

Your AI does the thinking on your own plan, so it is as quick and as thorough as the model you use.

  1. 1

    Open Settings → Connectors on claude.ai or in Claude Desktop.

  2. 2

    Choose Add custom connector, name it Idea Evaluator, and paste the address above.

  3. 3

    In a new chat, turn the connector on from the tools menu and write: “Evaluate my idea with the Idea Evaluator: …”

Other apps that support MCP connectors, such as Cursor or VS Code, use the same address. Menu names change between versions, and some apps offer connectors only on certain plans.

What a marked idea looks like

An example we wrote to show the format: the mark and the notes sit in the margin, beside the part of the idea they judge.

Example, not a real user's idea

A browser extension that turns recorded lectures into revision flashcards, with spaced-repetition reminders before exams.

For
Final-year project
Team
2 people
Time
3–6 months

Reality check

Building it is realistic; getting good flashcards out of messy lecture audio is the hard part. Transcription is good enough today, but turning a transcript into questions worth revising needs careful prompting and testing on real courses. Spaced repetition is a solved problem: use an existing algorithm.

Verdict

Build a narrower version. Start with one course you are taking and uploaded recordings only, and prove the cards are worth studying from before adding reminders.

Four pages, one plan

Each page builds on the one before it. You choose whether to turn the page, so you never plan an idea that failed the first question.

  1. 1

    Snapshot

    A feasibility score out of 10, your odds of finishing, difficulty, and an honest reality check.

  2. 2

    Summary

    Pros and cons, risks, what people are saying online and in the news, existing projects like yours, and an executive summary.

  3. 3

    Plan

    Scope, a suggested stack, phases with deliverables, the team you need, security, and a cost estimate.

  4. 4

    Hand-off

    Your idea judged as a whole against four questions, and a detailed requirements document with the user story map and wireframes, both as PDF or Word; the plan sent to Linear as a roadmap; and links to people who can help you build it.

Made for the moment you have to choose

  • Hackathon teams

    At kickoff, before anyone opens an editor.

    “Can we build something that works by the demo, or are we about to spend the night on the hard part?”

  • Final-year projects

    When you are choosing a topic or defending one to a supervisor.

    “Is this big enough to earn the marks, and small enough to finish by the deadline?”

  • Side projects

    Before you give it your weekends.

    “Does anyone besides me need this, and has someone already built it?”

Before it is marked, tell it what the project is for, your experience, and the time you have. The snapshot takes those into account.

The marking scheme

Every idea is marked against the same questions. A clever idea that nobody needs still loses marks.

Criteria

A real problem

Does it solve a clear problem that someone actually has?

Nobody you ask recognises the problem.

Worth solving

Is the pain big enough that people would change what they do today?

People shrug and keep their workaround.

Already done

Is it a common project, or does something like it already exist?

A free tool already does it well.

Something new

What does it do that the alternatives don't?

You can't name one difference.

Within reach

Can you build it with your skills and time, or does it need an expert?

The core part needs research you haven't done.

Someone wants it

Who are the users, and is there demand for it?

The only user you can name is you.

Questions before you start

Do I need an account?

Not here. Add the connector address to your AI and start a chat; the connector needs no sign-in. Some AI apps offer custom connectors only on certain plans.

Which AI apps does it work in?

Claude, ChatGPT (with developer mode on), the Gemini CLI, and any app that supports remote MCP connectors, such as Cursor or VS Code.

Where does my idea go?

Your own AI reads it and does the marking. The connector saves each stage your AI finishes for 90 days, at a link you can share, and runs the searches for similar projects and discussions. It runs no AI of its own.

How much should I trust the score?

Treat it as a second opinion from a strict examiner, not a guarantee. It is an AI's judgement, so check the risks it raises with real users and your supervisor.

Where do the similar projects come from?

From live searches of GitHub, Hacker News, Stack Exchange and Google News, run by the connector rather than remembered by the AI. Each one links to its source so you can read it yourself.

Find out before you spend the semester on it.

Your AI asks the time you have and your experience, and the idea is marked with that in mind. No account needed.

Add the connector