LaunchQ Build
Summary Email
Notion strategy board
Submit your idea
Webpage

LaunchQ — From Idea to Decision-Grade Market Intelligence in One Run

The Problem

Every founder has been there. You have an idea, you're excited, you spend weeks building, only to discover the market is saturated, the pricing is wrong, or three well-funded competitors already exist.

Market research is broken for early-stage founders. Hiring analysts costs thousands. Doing it manually takes weeks. Guessing costs you months of your life.

LaunchQ fixes this in less than a minute

What We Built

LaunchQ is an autonomous multi-agent market intelligence copilot built on Airia AI. A founder types their idea into a clean web interface and receives a complete, decision-grade market report — including a Go / Pivot / Avoid verdict backed by real-time web research.

The Report Includes:

✅ Ranked competitor analysis with threat scores
✅ Market saturation score (0-100)
✅ Trend velocity score (0-100)
✅ Customer pain-point clusters
✅ Pricing benchmark analysis
✅ Financial attractiveness estimate
✅ Strategic recommendation
✅ Go / Pivot / Avoid verdict
✅ Confidence score (0-100)

How We Built It

LaunchQ is a fully autonomous pipeline built inside Airia AI with the following architecture:

Input Form (HTML/JS)
      ↓
Amazon Nova Micro — parses and structures the idea
      ↓
Router — dispatches 3 parallel research agents
   ├── Brave Web Search — competitors and pricing
   ├── Exa Semantic Search — customer pain points and reviews  
   └── NewsAPI — market trends and signals
      ↓
Mistral Small — synthesises all data into a structured report
      ↓
Conditional Branch — checks confidence score
   ├── Route 1 (always) — DocGen creates Word document
   └── Route 2 (confidence ≥ 50)
         ├── Mistral Small Latest + Notion MCP — creates strategy board
         └── Mistral Small Latest + Gmail MCP — sends report to founder
      ↓
Output — download link, Notion board, email confirmation

The frontend is a custom HTML/JS single-page app with a cinematic intro sequence, served via Netlify and proxied through a Cloudflare Worker to handle CORS.

The Autonomous Workflow

What makes LaunchQ a true Track 2: Active Agents submission is the autonomous decision-making layer:

If confidence score ≥ 50 — LaunchQ automatically creates a Notion strategy board with action tasks, risk register, and competitor watchlist, then emails the full report to the founder
If confidence score < 50 — LaunchQ flags low data quality and only delivers the basic report, avoiding false confidence

The agent makes this decision itself. No human approval required.

The Stack

Component	Technology
Pipeline orchestration	Airia AI
Input parsing	Amazon Nova Micro
Web research	Brave Search, Exa, NewsAPI
Analysis and synthesis	Mistral Small Latest
Document generation	Airia DocGen
Strategy board	Mistral Small Latest + Notion MCP
Email delivery	Mistral Small Latest + Google Workspace MCP
Conditional routing	Airia Conditional Branch + Python
Frontend	HTML, CSS, JavaScript
Hosting	Netlify
CORS proxy	Cloudflare Workers

What We Learned

Multi-agent pipelines are harder than they look. The biggest challenge was not building individual agents — it was making them work together reliably. Variable persistence across steps, CORS handling between the frontend and Airia's API, and getting MCP tools to actually execute rather than just describe their actions were all real battles.

LLM routers are unreliable for numeric conditions. We initially used Airia's AI Router to branch on confidence score, but discovered the LLM would occasionally misread numbers. Replacing it with a Python Conditional Branch and a regex extractor made the routing 100% deterministic.

DocGen URL expiry is a real UX problem. Airia's generated document URLs expire in 5 minutes. We solved this by ensuring the email agent delivers the link immediately, so founders always have their report in their inbox before it expires.

Challenges We Faced

CORS — browsers block direct API calls to Airia. Solved with a Cloudflare Worker proxy
MCP tool calling — models sometimes describe what they would do instead of actually calling the tools. Solved with explicit tool-calling instructions in prompts
Variable persistence — Airia passes only the previous step's output downstream. Solved with a Python Code block that stores the founder's email early and makes it available to the email agent at the end
Dollar sign formatting — DocGen's template engine misread $ as LaTeX delimiters. Solved by instructing Mistral to write USD instead of $

What's Next

Google Drive integration — save reports permanently instead of 5-minute expiry links
Investor signal layer — detect whether VCs are actively funding this space
Multi-language support — run analysis for non-English markets
Saved history — let founders track all their ideas over time