AI-powered product viability analysis in about an hour, not weeks
Turn weeks of market research, competitive analysis, and strategic planning into a comprehensive AI-generated report suite. Quick validation in ~25 minutes, full VC-ready analysis in ~60-90 minutes.
Live Demo: venturepulse.shalusri.com
Experience the full VenturePulse workflow—upload your product spec, select AI models, and generate comprehensive viability reports. No installation required.
- Bring your own API key: You'll need an OpenRouter API key (free tier available)
- Public or Private: Choose to keep your analyses private or share them publicly for community feedback
VenturePulse is a prompt library + web application that generates McKinsey-quality product viability reports using AI. It analyzes your product idea across 19 critical dimensions (plus Provenance), organized into four strategic phases:
- Executive Summary - Viability scores, verdict, key highlights, recommended next steps
- Market Landscape - Competitors, timing, TAM/SAM/SOM, white space analysis
- User Stories - Core personas, jobs-to-be-done, problem scenarios
- Comparable Companies - Direct/indirect competitors, case studies, market positioning
- User Research - Research methodology, validation approach, interview guides
- Validation Experiments - Hypothesis testing, experiment design, success criteria
- Technical Feasibility - Architecture, complexity, AI/low-code implementation, risks
- Competitive Advantage - Moats, defensibility, competitive scoring matrix
- Business Model - Pricing, unit economics, financial projections
- Legal & Compliance - Regulatory requirements, IP considerations, privacy
- MVP Roadmap - Feature prioritization matrix, phased timeline, implementation strategy
- Customer Journey - Acquisition to advocacy lifecycle, touchpoints
- Go-to-Market - ICP analysis, distribution channels, acquisition strategy
- Partnerships - Strategic alliances, integration opportunities, ecosystem
- Expansion Plan - Geographic/vertical growth strategy, market entry
- Success Metrics - KPIs across technical/engagement/business dimensions, risk register
- Funding Strategy - Capital requirements, investor narrative, fundraising roadmap
- Exit Strategy - Acquisition targets, exit timeline, valuation drivers
- Pitch Narrative - Compelling story, key messages, presentation framework
- Provenance - Analysis transparency, model details, generation timestamp
The output: Up to 20 beautifully formatted HTML reports—each independently comprehensive and presentation-ready. Choose Quick Analysis (7 core sections) or Full Analysis (all 19 sections).
- ❌ 30+ fragmented AI conversations
- ❌ Manual research across competitors, pricing, tech stack
- ❌ Scattered insights in notes, docs, spreadsheets
- ❌ Missing critical dimensions (compliance, risks, defensibility)
- ❌ No structured decision framework
- ✅ One command generates up to 19 comprehensive reports
- ✅ Quick Analysis (7 sections) or Full Analysis (19 sections)
- ✅ Automated competitive research and market analysis
- ✅ Structured scoring across all viability dimensions
- ✅ Professional formatting ready for stakeholders/investors
- ✅ Costs $0 (free models) to $10 (premium full analysis)
What used to cost thousands in consulting fees now costs the price of a coffee.
- OpenRouter API key (free tier available)
- Docker and Docker Compose
# Clone the repository
git clone https://github.com/knightsri/VenturePulse.git
cd VenturePulse
# Create .env with your configuration (see .env.example)
cp .env.example .env
# Edit .env and configure:
# - OPENROUTER_API_KEY (required)
# - OAuth credentials for Google/GitHub login (optional)
# - PORT (default: 8501)
# Run with Docker Compose
docker-compose up
# Open http://localhost:8501 (or custom PORT from .env)
# Sign in with Google, GitHub, or Dev mode
# Enter your OpenRouter API key in SettingsTime estimates:
- Quick Analysis (7 sections): ~25 minutes
- Full Analysis (19 sections): ~60-90 minutes
The Docker-based web application is the recommended way to use VenturePulse.
- 📤 Upload specs - Drag & drop or paste your project specification
- 🤖 Multi-model analysis - Run the same spec through multiple AI models
- ⚡ Parallel section generation - Generate all sections simultaneously for faster analysis
- 📊 Real-time progress - See elapsed time, cost, and retry status
- 🔄 Automatic retries - Failed sections retry with exponential backoff
- 💰 Cost tracking - See per-section and total costs in Provenance
- 🔬 Compare results - Side-by-side comparison of outputs from different models
- 📚 Analysis history - Browse and view past analyses
- 🔐 OAuth Authentication - Sign in with Google, GitHub, or Dev mode
- 👥 Multi-user support - Each user has their own projects and analyses
- 🌍 Public/Private projects - Share analyses publicly or keep them private
- 🔑 Per-user API keys - Each user enters their own OpenRouter API key
- 👍 Section feedback - Rate analysis quality with thumbs up/down (coming soon)
- 📊 Enhanced comparison - Charts, metrics, and recommendations (coming soon)
The web app lets you:
- Select multiple models for analysis
- Compare results with side-by-side section comparison
- View metrics - timing, cost, and quality indicators per model
Legacy Versions: Standalone Streamlit app and CLI scripts are archived in
Archive/for reference. These may not work with current prompts.
- Sign in - Use Google, GitHub, or Dev mode
- Add your API key - Go to Settings and enter your OpenRouter API key
- Create a project - Upload or paste your product specification
- Choose models - Select one or more AI models to analyze with
- Run analysis - Choose Quick (7 sections) or Full (19 sections)
- Review results - Browse sections, compare models, export reports
VenturePulse works with 100+ models via OpenRouter. These are the tested, reliable models available in the UI:
| Model | Speed | Quality | Cost (Full 19) | Best For |
|---|---|---|---|---|
anthropic/claude-sonnet-4 |
⚡ Slower | ✅✅✅ Best | 💰💰💰 ~$5-10 | Investor-ready analysis |
anthropic/claude-3.5-sonnet |
⚡ Slower | ✅✅✅ Excellent | 💰💰💰 ~$4-8 | Detailed strategic analysis |
openai/gpt-4o |
⚡⚡ Medium | ✅✅ Very Good | 💰💰 ~$3-6 | Balanced speed & quality |
openai/gpt-4o-mini |
⚡⚡⚡ Fast | ✅ Good | 💰 ~$0.30-0.60 | Quick validation |
google/gemini-2.5-pro |
⚡⚡ Medium | ✅✅ Excellent | 💰💰 ~$3-5 | Solid validation |
deepseek/deepseek-chat |
⚡⚡⚡ Fast | ✅ Good | 💰 ~$0.30-0.50 | Budget-friendly testing |
deepseek/deepseek-v3.2 |
⚡⚡ Medium | ✅✅ Very Good | 💰 ~$0.50-1.00 | GPT-5 class reasoning |
x-ai/grok-4.1-fast |
⚡⚡⚡ Fast | ✅✅ Good | 💰 FREE | 2M context, agentic |
x-ai/grok-4-fast |
⚡⚡⚡ Fast | ✅✅ Good | 💰💰 ~$1-3 | 2M context, multimodal |
qwen/qwen3-max |
⚡⚡ Medium | ✅✅ Good | 💰 ~$0.50-1.00 | 256K context, multilingual |
qwen/qwen-2.5-72b-instruct |
⚡⚡ Medium | ✅ Good | 💰 ~$0.30-0.60 | Structured output |
z-ai/glm-4.7 |
⚡⚡ Medium | ✅✅ Good | 💰 ~$0.50-1.00 | 203K context, coding |
z-ai/glm-4.5-air |
⚡⚡⚡ Fast | ✅ Good | 💰 ~$0.20-0.40 | Budget-friendly |
mistralai/mistral-large |
⚡⚡ Medium | ✅✅ Good | 💰💰 ~$2-4 | European alternative |
meta-llama/llama-3.3-70b-instruct |
⚡⚡ Medium | ✅ Good | 💰 ~$0.50-1.00 | Open source option |
My recommendation: Start with GPT-4o-mini, DeepSeek, or Grok-4.1-fast (free!) for quick validation. For investor-ready analysis, use Claude Sonnet 4 or Gemini 2.5 Pro—the depth is worth every penny.
See all models: https://openrouter.ai/models
VenturePulse/
├── app/
│ └── main.py # FastAPI application entry point
├── prompts/
│ ├── common-instructions.md # Shared analysis guidelines
│ └── sections/
│ ├── section01-executive-summary.md
│ ├── section02-market-landscape.md
│ └── ... (19 total section prompts)
├── data/ # User data (gitignored)
│ └── reports/ # Generated analysis reports
├── Archive/ # Legacy versions (CLI, Streamlit)
├── examples/
│ └── sample-project/
│ └── smartplate-idea.md # Sample project description
├── Dockerfile # Docker build config
├── docker-compose.yml # Docker Compose config
├── requirements.txt # Python dependencies
├── CLAUDE.md # Claude Code guidance
└── README.md
Current approach (v2.0): Sequential specialized prompts → 19 focused HTML reports + Provenance
- ✅ Each section uses tailored prompts for that analysis type
- ✅ Web UI with multi-model comparison and cost tracking
- ✅ Quick (7 sections) or Full (19 sections) analysis modes
- ✅ Easy to regenerate any individual section
- ✅ Better handling of complex projects
-
You provide: A markdown/text file describing your product idea (1-3 pages ideal)
-
VenturePulse orchestrates:
- Loads common analysis instructions
- Sequentially generates up to 19 specialized reports:
- Each section calls OpenRouter API with your chosen model
- Later sections reference insights from earlier reports
- Each generates its own styled HTML file
- Creates provenance metadata for transparency
-
You get: all comprehensive HTML reports in a timestamped folder
Total time: 25-90 minutes depending on analysis depth (varies by model) Total cost: $0 (free models) to $5-15 (premium full analysis)
Each analysis creates separate HTML files:
data/reports/{user_id}/{project_slug}/{analysis_id}/
├── section01-executive-summary.html
├── section02-market-landscape.html
├── section03-user-stories.html
├── ... (up to 19 sections)
├── section19-pitch-narrative.html
├── section20-provenance.html
├── project-spec.md
└── metadata.json
- Professional Design: Executive-ready styling, modern UI
- Standalone Reports: Each HTML works independently—easy to share specific sections
- Rich Visualizations: Scoring matrices, competitive comparisons, feature prioritization grids
- Comprehensive Coverage: 4,000-8,000 total words across all dimensions
- Portable: No external dependencies, works offline, easy to email/share
- Print-Ready: Professional formatting for hard copies
Executive Summary gives you:
- Clear GO BUILD / PROTOTYPE FIRST / RE-VALIDATE verdict
- Top 3 highlights of your idea
- Viability scores across 5 dimensions (1-10 scale)
- Critical success factors and key risks
- Recommended next steps
Market Landscape includes:
- 3-5 existing competitors with detailed analysis
- Competitive scoring matrix (your idea vs. alternatives)
- Market timing rationale ("Why now?")
- White space identification
- TAM/SAM/SOM estimates
Success Metrics provides:
- Specific KPIs across technical, engagement, and business dimensions
- Example: "99.5% system uptime, 45% 30-day retention, 4.7x LTV:CAC ratio"
- Risk register with probability, impact, and mitigations
- Comprehensive scoring with gap analysis
...and 16 more equally detailed reports covering user stories, legal compliance, customer journey, partnerships, funding strategy, and more.
Your project description can be simple or comprehensive. Here's what works best:
- Project name and one-sentence description
- Problem you're solving
- Target audience
- Proposed solution
- Business model ideas or pricing thoughts
- Technical approach (if you have preferences)
- Market opportunity (if known)
- Key competitors you're aware of
- Your background/resources/constraints
- Any specific concerns or risks
Example: See examples/sample-project/smartplate-idea.md
Format: Markdown (.md), plain text (.txt), or PDF
Length: 1-3 pages is ideal, but more is fine—the tool handles it
Since releasing VenturePulse, it's been used for:
"Should I spend my weekends building this?"
- Get clear go/no-go decision in under 2 hours vs. 2-3 weeks of research
"We're considering building an application of -so-and-so- domain — does it work?"
- Run comparative analyses of competition and feasibility
"A client wants us to build X—is it viable?"
- Generate feasibility report before committing resources
"We're looking at investing in a potential startup—what are the real risks?"
- Independent AI-powered analysis of the opportunity
"I want to understand product strategy better"
- Study the framework and analysis approach
| Variable | Default | Description |
|---|---|---|
OPENROUTER_API_KEY |
(required) | Your OpenRouter API key |
PORT |
8080 |
Application port |
DEFAULT_MODEL |
anthropic/claude-sonnet-4 |
Pre-selected model in UI |
MAXRETRY |
3 |
Maximum retry attempts for failed API calls |
MAX_PARALLEL_SECTIONS |
10 |
Maximum concurrent section generation workers |
Example .env file:
OPENROUTER_API_KEY=sk-or-v1-...
PORT=8888 # Run on custom port
DEFAULT_MODEL=openai/gpt-4o # Change default model
MAXRETRY=5 # Retry up to 5 times
MAX_PARALLEL_SECTIONS=5 # Use up to 5 parallel workersVenturePulse automatically retries failed API calls with exponential backoff + jitter:
- Retryable errors: Rate limits, timeouts, server errors (502/503/504)
- Non-retryable errors: Invalid API key, model not found, content policy violations
- Backoff formula:
base_delay * (2 ^ attempt) + random(0, jitter_max)- Example: attempt 1 = ~2-4s, attempt 2 = ~4-6s, attempt 3 = ~8-10s
Failed sections after max retries are marked in the Provenance report with error details.
Enable the "Parallel Section Generation" toggle in the UI to generate all sections simultaneously:
- Faster: Completes Full Analysis in ~15-25 minutes vs. ~60-90 minutes sequential
- Trade-off: Uses more API quota simultaneously (may hit rate limits on free tiers)
- Best for: Premium models with higher rate limits
Set DEFAULT_MODEL in your .env file:
DEFAULT_MODEL=openai/gpt-4oEdit prompt files in prompts/sections/:
- Modify existing section prompts
- Add domain-specific questions
- Adjust scoring criteria
- Change output format
This is encouraged! Fork and customize for your industry/domain.
Each section has optimized temperature and seed settings to balance accuracy vs. creativity:
| Temperature | Purpose | Sections |
|---|---|---|
| 0.2 (Precision) | Factual data, no hallucination | Executive Summary, Market Landscape, Comparable Companies, Technical Feasibility, Business Model, Legal & Compliance, Success Metrics, Funding Strategy, Exit Strategy |
| 0.3 (Grounded) | Real methodologies, realistic planning | User Research, Validation Experiments, Competitive Advantage, MVP Roadmap, Partnerships |
| 0.5 (Balanced) | Creative but realistic strategies | Customer Journey, Go-to-Market, Expansion Plan |
| 0.7 (Creative) | Personas, storytelling, ideation | User Stories, Pitch Narrative |
Seed for reproducibility: Precision sections (0.2) use a fixed seed (42) so running the same analysis twice produces consistent results. This is critical for market data, competitor analysis, and financial projections.
To modify these settings, edit the SECTIONS configuration in app/routes/analysis.py.
# Check if Docker is running
docker info
# Rebuild containers
docker-compose down
docker-compose build --no-cache
docker-compose up- Dev mode: Ensure
DEV_MODE=truein.envfor local development - Production: Verify OAuth credentials (GOOGLE_CLIENT_ID, etc.) are correct
- Callback URL: Ensure OAuth app callback URLs match your deployment
- Check model name is exact (case-sensitive)
- Visit https://openrouter.ai/models to verify
- Some models require special access or credits
- Check OpenRouter credits: Visit https://openrouter.ai/credits
- Try different model: Some models have rate limits
- Check logs:
docker-compose logs -f
- Open in modern browser (Chrome, Firefox, Safari, Edge)
- Check if HTML file is complete (not truncated)
- Try regenerating the section
- Expected: ~25 minutes for Quick Analysis (7 sections), ~60-90 minutes for Full Analysis (19 sections)
- Premium models (Claude) are slower but higher quality
- Try faster model:
google/gemini-2.0-flash-exp:free
- ✅ Expanded to 19 specialized sections (from 8)
- ✅ Web UI with multi-model comparison
- ✅ Quick (7 sections) / Full (19 sections) / Custom analysis modes
- ✅ Sequential and parallel model execution
- ✅ Parallel section generation - All sections generated simultaneously
- ✅ Automatic retry logic - Exponential backoff with jitter for failed API calls
- ✅ Cost tracking - Per-section and total costs in Provenance
- ✅ Failure handling - Graceful handling of partial failures with detailed error reporting
- ✅ Grouped section navigation (Foundation/Strategy/Execution/Future)
- ✅ Docker deployment support
- 🔄 Ollama/local model support
- 🔄 Ideas library for saved projects
- 🔄 Report export (PDF, combined HTML)
Contributions welcome! Here's where we need help:
- Prompt Engineering: Improve section prompts for better insights
- Industry Templates: Create specialized prompts (fintech, healthcare, B2B SaaS, etc.)
- Model Testing: Test different models and report quality/cost findings
- Documentation: Improve guides, add tutorials, create videos
- Bug Reports: Find issues, suggest improvements
- Feature Requests: What would make this more useful?
- Example Projects: Contribute sample analyses
- Translations: Internationalize prompts and docs
Just open an issue or submit a PR on GitHub!
MIT License - see LICENSE for details.
TL;DR: Use it, modify it, distribute it, commercialize it freely.
The generated reports are yours—use them however you want.
- Built with OpenRouter for unified multi-model AI access
- Inspired by months of real-world product validation experience
- Prompt refinement informed by 50+ test analyses
- Thanks to the open-source community
Special thanks to early testers and contributors.
- Live Demo: https://venturepulse.shalusri.com
- GitHub: https://github.com/knightsri/VenturePulse
- Issues/Bugs: https://github.com/knightsri/VenturePulse/issues
- Discussions: https://github.com/knightsri/VenturePulse/discussions
- OpenRouter: https://openrouter.ai
- Model Directory: https://openrouter.ai/models
- Author Blog: https://shalusri.com
Depends on the model and analysis depth:
| Model | Quick (7 sections) | Full (19 sections) |
|---|---|---|
deepseek/deepseek-chat |
~$0.10-0.20 | ~$0.30-0.50 |
openai/gpt-4o-mini |
~$0.10-0.25 | ~$0.30-0.60 |
meta-llama/llama-3.3-70b-instruct |
~$0.20-0.40 | ~$0.50-1.00 |
mistralai/mistral-large |
~$0.80-1.50 | ~$2-4 |
google/gemini-2.5-pro |
~$1-2 | ~$3-5 |
openai/gpt-4o |
~$1-2 | ~$3-6 |
anthropic/claude-3.5-sonnet |
~$1.50-3 | ~$4-8 |
anthropic/claude-sonnet-4 |
~$2-4 | ~$5-10 |
My workflow: Start with DeepSeek or GPT-4o-mini to filter obviously flawed ideas quickly. For promising ideas, use Claude Sonnet 4 or Gemini 2.5 Pro for investor-ready analysis. The Web UI shows exact cost per section and total in the Provenance report.
- Quick Analysis: ~25 minutes (7 core sections)
- Full Analysis: ~60-90 minutes (all 19 sections)
- Premium models are slower but produce higher quality analysis
Time varies by model—faster models complete quicker, premium models are slower but produce better analysis.
Yes! MIT license allows commercial use. The generated reports are yours to use however you want—pitch investors, share with clients, include in proposals, etc.
VenturePulse provides strategic insights based on AI reasoning and training data. It's exceptionally good at:
- ✅ Identifying competitive landscape
- ✅ Spotting risks you might miss
- ✅ Suggesting pricing strategies
- ✅ Structuring your thinking
However, always:
- Validate with domain experts
- Conduct your own customer research
- Verify competitive intelligence with primary sources
- Perform financial due diligence
Think of it as: A brilliant research assistant and strategic advisor, not a replacement for human judgment. It accelerates your thinking by 10x, but you still need to validate the insights.
Using ChatGPT/Claude directly:
- Generic, surface-level analysis
- Inconsistent structure across conversations
- You have to remember what to ask
- Missing critical dimensions
- No comparative framework
- Takes 2-3 weeks of iterating
Using VenturePulse:
- Comprehensive, structured analysis across 19 dimensions
- Proven framework covering all viability aspects
- Automatic competitive research
- Professional formatting with cost tracking
- Takes 25-90 minutes (vs. weeks manually)
The prompts took months to refine. You're getting battle-tested analysis templates that have been used on 50+ real projects.
Absolutely! That's the whole point of open-source. You can:
- Edit existing section prompts in
prompts/sections/ - Add new sections (e.g., regulatory deep-dive for healthcare)
- Remove sections you don't need
- Adjust scoring criteria
- Change output format
Fork it and make it your own. Share improvements back with the community!
Strategic decision based on real-world use:
Advantages of separate files:
- ✅ Share specific sections with different stakeholders
- ✅ Email just the "Executive Summary" to busy executives
- ✅ Regenerate individual sections without re-running everything
- ✅ Avoid token limit issues with complex projects
- ✅ Better quality—each section gets full AI focus
Previous version (one big file with tabs) had truncation issues and less detailed analysis.
Future version might include an optional "combined view" while keeping individual files.
Your data never leaves your machine except for the API call to OpenRouter.
- API calls are encrypted (HTTPS)
- OpenRouter doesn't train on your data (per their policy)
- Generated reports are saved locally on your machine
- No telemetry, no tracking, no data collection by VenturePulse
For extra security:
- Use a self-hosted LLM (future feature)
- Review and redact your project file before analysis
- Check OpenRouter's privacy policy
Not currently, but maybe in a future release:
- Ollama integration for fully local analysis
- LM Studio support
- Self-hosted model options
For now, you need internet + OpenRouter API access.
OpenRouter provides:
- ✅ Single API key for 100+ models
- ✅ Unified pricing and billing
- ✅ Fallback routing if model is down
- ✅ Automatic load balancing
- ✅ Easy model switching
- ✅ Usage analytics
You could modify the scripts to call OpenAI/Anthropic/Google directly, but OpenRouter makes multi-model testing much easier.
AI analysis has limitations:
- May miss recent market developments (post-training cutoff)
- Can't validate assumptions with real users
- Doesn't have your domain expertise
- May not know niche competitors
How to use VenturePulse effectively:
- Start: Generate initial analysis
- Validate: Check competitive research, verify claims
- Augment: Add your domain knowledge and corrections
- Iterate: Regenerate specific sections with more context
- Decide: Use as input to your decision, not the sole factor
It's a strategic thinking tool, not a crystal ball.
- ✅ Solo founders validating side project ideas
- ✅ Startup teams exploring pivot opportunities
- ✅ Product managers assessing new feature viability
- ✅ Consultants scoping client projects
- ✅ Investors conducting preliminary due diligence
- ✅ Students learning product strategy frameworks
- ✅ Agencies evaluating build vs. buy decisions
# 1. Clone the repo
git clone https://github.com/knightsri/VenturePulse.git
cd VenturePulse
# 2. Configure
cp .env.example .env
# Edit .env with your OPENROUTER_API_KEY
# 3. Run with Docker
docker-compose up
# 4. Open http://localhost:8501 and sign inStop spending weeks on viability analysis. Start building products faster.
Built by founders, for founders 🚀
Built in public 🌍
Built with AI 🤖
Questions? Open an issue or discussion.
