You don't need to guess at a user's real pain points — the reviews, the complaints and the search data are already public, and a stack of free, open-source AI tools can pull them for you. Several are MCPs: install one into Claude or Cursor and just ask in plain language.
📱 Step 1 — App Store / Play Store Reviews
These three pull the reviews for you:
- mcp-appstoreMCP
An open-source MCP that searches both app stores, pulls reviews and analyses sentiment and themes — the best fit for a live demo.
MIT-licensed, about 1k stars, pulls Play Store reviews.
The Node version does the same for Apple. Skip the older Python scraper — it was archived in 2021.
💬 Step 2 — Reddit Complaints
For complaints, go to Reddit:
Searches subreddits and posts, linking back to every source. The hosted version needs no API key, but its licence is unconfirmed — worth checking before you rely on it.
📈 Step 3 — Search Demand
To check whether a demand is big enough, ask the AI to compare search trends:
Ask the AI to compare search terms and it pulls the Google Trends chart for you — the same thing already on the slide.
📑 Step 4 (Advanced) — Full Research Reports
For a fully cited market-research report, these two are the advanced option — both need their own API keys:
- GPT ResearcherApache-2.0
Apache-2.0, 29.5k stars, writes fully cited market-research reports. Needs your own OpenAI and Tavily API keys — treat it as the advanced take-home option.
A similar tool from LangChain, worth trying if you want a second option.
Still manual, for now: No solid open-source tool was found for pulling revenue or complaint posts from X (Twitter) — this step still has to be done by hand.
🧩 Step 5 — Turn Raw Data Into a Pain-Point List
Once you have reviews, posts or Trends data, paste them into Claude with this prompt:
We just pulled a batch of app reviews / Reddit posts / Google Trends data using the tools above — pasted below. Please help me: 1. Group the recurring user pain points, and quote the most common original wording for each. 2. Point out which pain point the market still hasn't solved well. 3. If we built one AI feature to solve the top pain point, how should it be designed? (Paste the reviews / posts / data you pulled)
These tools don't invent the idea for you — they just get you real user evidence faster than anyone else. Enough evidence is what tells you which pain point is actually worth building for. Only once that is solid should you think about how AI solves it.
