FoundryScan

Open-source micro-SaaS opportunity discovery platform.

FoundryScan turns public discussions and trend signals into ranked micro-SaaS opportunities developers can evaluate and build.

No API keys required for most sources  ·  No SaaS subscriptions  ·  Runs entirely on your machine with Docker

---

---

How it works

1. Configure a scan - enter keywords describing the space you want to explore (e.g. "manual process", "excel nightmare", "wish there was a tool")
2. Collect - the backend retrieves Reddit posts and comments, Hacker News "Ask HN" threads, Google Trends data, and Product Hunt discussions in parallel
3. Build a prompt - collected raw data is assembled into a structured prompt optimized for LLM analysis
4. Paste to your LLM - copy the prompt, paste it into ChatGPT, Claude, Gemini, or any LLM of your choice, then paste the response back
5. Review opportunities - the app parses the LLM output into structured opportunity cards with scores, evidence, MVP features, and build estimates
6. Track and export - set statuses (saved, evaluating, building…), add personal notes, download individual opportunities as .md or .txt

---

Features

• Multi-source collection: Reddit, Hacker News, Google Trends, Product Hunt
• Playwright fallback for JavaScript-rendered pages and bot-protection bypass
• Structured scoring: pain intensity, trend momentum, competition gap, MVP feasibility
• Infinite-scroll opportunity list with URL-persisted filters (status, sort, min score)
• Right-click / long-press context menu on opportunity cards (change status, download)
• Scan history with archive/delete and Active / Archived / All tabs
• Dark/light theme (RawBlock design system - no rounded corners, no shadows)
• Responsive mobile layout with slide-in navigation drawer
• Export: scan report as Markdown, individual opportunities as .md or .txt
• 10-minute hard timeout on collection, 30-day auto-cleanup of raw data
• Fully self-hosted - no external services required beyond an LLM of your choice

---

Tech stack

Layer	Technology
Frontend	Next.js 15 (App Router), React 19, Tailwind CSS v4
Backend	FastAPI, Python 3.12, Pydantic v2
Database	PocketBase (embedded SQLite, built-in migrations)
Collection	HTTPX, Playwright (headless Chromium fallback)
Trends	PyTrends
Container	Docker Compose

---

Quick start

Prerequisites

• Docker with Compose v2
• ~1.5 GB disk space (Playwright Chromium + Python + Node deps)

1. Clone

git clone https://github.com/daiv05/foundry-scan.git
cd foundry-scan

2. Configure environment

cp backend/.env.example  backend/.env
cp frontend/.env.example  frontend/.env

Edit backend/.env - the only value you must change is PB_ADMIN_PASSWORD. Everything else works out of the box.

Optionally add a PRODUCTHUNT_TOKEN (free at api.producthunt.com).

3. Build and run

docker compose up --build

First build takes 3–5 minutes. Subsequent starts are fast.

Service	URL
Frontend	http://localhost:7110
Backend API	http://localhost:7120/docs
PocketBase admin	http://localhost:7130/_/

4. Launch your first scan

1. Open http://localhost:7110
2. Click New Scan and enter keywords for your target market
3. Wait for collection to finish (~1–3 minutes)
4. Copy the generated prompt and paste it into your LLM
5. Copy the LLM response, go back to the scan page, and paste it in
6. Your ranked opportunities appear immediately

---

Project structure

foundry-scan/
├── backend/
│   ├── app/
│   │   ├── routers/        HTTP endpoints (scans, opportunities, settings, configs)
│   │   ├── services/
│   │   │   ├── collector/  Per-source collectors (reddit, hn, trends, producthunt)
│   │   │   ├── scan_service.py     Orchestrates the full pipeline
│   │   │   ├── processor.py        Deduplicates and filters raw signals
│   │   │   ├── prompt_builder.py   Assembles the LLM prompt
│   │   │   ├── response_parser.py  Parses and persists LLM output
│   │   │   └── cleanup_service.py  Periodic raw-data expiry
│   │   ├── models/         Pydantic schemas
│   │   └── db.py           PocketBase async client
│   └── Dockerfile
├── frontend/
│   ├── src/
│   │   ├── app/            Next.js App Router pages
│   │   ├── components/     UI components (RawBlock design system)
│   │   ├── lib/            API client, types, export utilities
│   │   └── hooks/          useLongPress, etc.
│   └── Dockerfile
├── pocketbase/
│   ├── pb_migrations/      Schema migrations (auto-applied on start)
│   └── Dockerfile
├── docs/
│   ├── SPEC.md             Full product specification
│   └── features/           Per-feature implementation notes
└── docker-compose.yml

---

Adding a new data source

1. Create backend/app/services/collector/your_source.py with an async collect(keywords) function that returns list[dict]
2. Register it in scan_service.py alongside the other collectors
3. Optionally add a config toggle in backend/app/config.py

See docs/features/F04-default-collector.md for a reference implementation.

---

Documentation

Document	Contents
LOCAL-SETUP.md	Full env variable reference, hot-reload dev workflow, rebuilding tips
FAQ.md	Why no LLM API, PocketBase rationale, common questions, dev tips
CONTRIBUTING.md	Code style, adding collectors, PR process
LEGAL.md	Data collection disclaimer, third-party ToS notes, responsible use
docs/SPEC.md	Full product specification

---

Contributing

See CONTRIBUTING.md.

---

License

MIT - free to use, modify, and distribute.

Legal

FoundryScan collects publicly available data from third-party platforms for personal market research purposes. Users are responsible for complying with the terms of service of any platform from which data is collected. See LEGAL.md for full details.