Market Scout AI

Validate any startup idea in 60 seconds using live competitor data. Powered by Perplexity + Claude — built for non-tech founders who refuse to build products nobody wants.

Comment

Inspiration

90% of startups fail because founders build products nobody wants. Not because they lack talent — because they skip validation. Market research tools either cost thousands of dollars, take weeks, or require a data analyst to interpret.

We asked: what if a non-technical founder could type a raw idea and get real, actionable market intelligence in under 60 seconds?

That question became Market Scout AI.

What It Does

Market Scout AI is a real-time market validation engine. A founder types any startup idea — raw, messy, unpolished — and the platform immediately returns:

Output Description
Top Competitors 3 real companies with live URLs, sourced from the web in real time
Market Weaknesses The 5 most common pain points customers have with existing solutions
Your Positioning An AI-generated H1 headline + subheadline ready to use
Verdict GO / CAUTION / NO-GO with a clear strategic rationale
What to Build First 5 prioritized MVP feature recommendations

No jargon. No 40-page reports. Just a clean, structured output a founder can act on immediately.

How We Built It

Market Scout uses a two-stage AI pipeline running entirely in real time on the frontend:

Stage 1 — Live Web Intelligence (Perplexity API)

The user's idea triggers a live web search via the Perplexity sonar model. This returns grounded, real-time competitor data — not hallucinated company names, but actual URLs with actual customer complaints sourced from the live web.

Stage 2 — Strategic Analysis (Anthropic Claude)

The raw Perplexity output is passed directly to Claude (claude-sonnet-4-5) with a structured prompt. Claude transforms the competitive intelligence into actionable positioning: headline copy, market weaknesses, MVP priorities, and a GO/CAUTION/NO-GO verdict with rationale.

Stage 3 — Silent Automation (Make.com Webhook)

After results display, a background webhook fires to Make.com — logging the lead (name, email, idea, verdict) for CRM follow-up. Zero friction for the user.

User Input → Perplexity API (live search) → Claude API (analysis) → Structured Output ↓

Make.com Webhook (background)

Tech Stack:

Layer Technology
Frontend Lovable (React + TypeScript)
Live Search Perplexity API (sonar model)
AI Analysis Anthropic Claude (claude-sonnet-4-5)
Automation Make.com Custom Webhook
Hosting Lovable (production)
Secrets Lovable server-side environment variables (never exposed client-side)

Challenges We Ran Into

1. Model String Errors

The initial Claude API call returned a 404 because Lovable generated an incorrect model string (claude-sonnet-4-20250514). Diagnosed and corrected to claude-sonnet-4-5 — verified against the Anthropic API directly.

2. API Key Exposure Risk

Early builds risked exposing API keys client-side. Solved by routing all API calls through Lovable server-side environment variables — neither the Perplexity nor Anthropic key is ever visible in the browser.

3. Latency UX

The two-stage API pipeline takes 10–15 seconds. Without feedback, users abandoned. Solved with a terminal-style animated loader showing which stage is running in real time ("Perplexity scanning..." → "Claude analyzing...").

4. JSON Parsing Reliability

Claude occasionally returned markdown-wrapped JSON, breaking the parser. Solved by explicitly instructing Claude to return strict JSON with no preamble, and adding a JSON.parse fallback with replace(/json|/g, '').

Accomplishments We're Proud Of

  • ✅ Full end-to-end pipeline live — Perplexity → Claude → structured output → Make.com webhook, all working in production
  • ✅ Real competitor URLs — not hallucinated names, but actual verified companies sourced from live web data
  • ✅ Sub-15 second total latency — both API calls complete and render before a user loses patience
  • ✅ Zero technical knowledge required — tested with a 60-year-old non-technical founder who completed a full analysis in 90 seconds
  • ✅ Built and shipped in under 72 hours — from blank Lovable canvas to live production URL

What We Learned

Perplexity and Claude are genuinely complementary, not redundant. Perplexity is a retrieval engine — fast, grounded, factual. Claude is a reasoning engine — structured, strategic, copywriting-ready. Using each for what it does best produces output neither could achieve alone.

We also learned that the biggest UX challenge in AI tools isn't the AI — it's managing the latency gap. The animated terminal loader reduced perceived wait time more than any performance optimization we made.

What's Next for Market Scout AI

Q3 2026 — Integration Layer

  • Connect validated ideas directly to Medusa Black Labs' 72h MVP build service
  • Add PDF export of the full market analysis report
  • Integrate Notion output (one-click save to founder's workspace)

Q4 2026 — Depth Mode

  • Add TAM/SAM/SOM estimation layer
  • Competitive pricing analysis (scrape pricing pages of identified competitors)
  • Historical trend data overlay (is this market growing or shrinking?)

2027 — Platform

  • API access for accelerators and startup studios
  • Bulk analysis mode for investors evaluating deal flow
  • White-label version for startup incubators

Market Scout AI is the entry point. The validated idea becomes the brief. The brief becomes the MVP. That's the Medusa Black Labs flywheel.

Built With

Share this project:

Updates