How AngelCheck Works
Privacy-first AI analysis powered by pattern recognition from 25 years of venture outcomes. Evidence-based scoring with multi-layer quality assurance.
Three Ways to Use AngelCheck
Chrome Extension
One-click capture on AngelList
- • Analyze while you browse
- • Auto-extracts deal memos
- • Queued for background analysis
Works on any deal page
Full Analysis
Deep 8-criteria scoring in 2-5 minutes
- • Evidence-based scores
- • Red flags and risk analysis
- • VC frameworks and exit math
Free: 10 analyses/month
Second Opinion
Claude Opus 4 critical review
- • Checks for missed red flags
- • Score calibration validation
- • Different AI perspective
Pro: 10 free/month ($2.99 each otherwise)
Compare Deals Side-by-Side
Evaluating multiple investment opportunities at once? The comparison feature helps you make informed allocation decisions by showing all your analyzed deals in one view.
What You Can Compare
- • Overall scores and recommendations
- • All 8 investment criteria side-by-side
- • Risk tiers and investment stages
- • Red flags and concerns
- • Exit potential and return scenarios
- • Position sizing recommendations
AI Chat on Comparisons
The AI chat assistant is available when comparing deals. Ask questions like:
- • "Which deal has better founder-market fit?"
- • "What are the key risks I should weigh?"
- • "How should I allocate $100k between these 3?"
- • "Which has the clearest path to Series A?"
Pro Tip: Save your analyzed deals to your library to easily pull them into comparison view. You can compare up to 10 deals at once.
Unlike uploading memos to Claude individually, AngelCheck maintains structured analysis that you can compare side-by-side, track over time, and reference with AI chat that remembers full context.
What Makes AngelCheck Different from Using Claude Myself?
While Claude Sonnet 4.5 is excellent, AngelCheck provides a complete investment analysis system:
Privacy-First Local Anonymization
Sensitive data scrubbed on your device before AI processing
Custom Prompts with Venture Pattern Recognition
Trained on 25 years of unicorn successes and failures
Multi-Layer Quality Assurance
Accuracy check, auto-retry, and polish to catch hallucinations
Consistent 8-Criteria Framework
Structured scoring for apples-to-apples comparison
Saved Library + Comparison Tools
Track deals over time, compare multiple opportunities
Financial Calculators + Position Sizing
Equity math, dilution scenarios, exit value modeling
The Analysis Process
Paste Your Deal Info
Copy the memo from AngelList, Republic, syndicate emails - no special formatting needed. Optionally add pitch deck URLs or video links (YouTube, Loom).
Privacy-First Processing
Before any AI analysis begins, we anonymize sensitive information locally on your device:
- • Company names removed
- • Founder names removed
- • Locations removed
- • Financial details scrubbed
OpenAI never sees the raw sensitive data. Only pre-anonymized text is sent for analysis.
AI Analysis Powered by Claude (2-5 Minutes)
We use Claude Sonnet 4.5 by Anthropic for 100% of investment analysis. Claude analyzes the deal across 8 criteria, using pattern recognition from historical venture outcomes. Claude Opus 4 is also available as a Second Opinion for critical validation.
Why not just use Claude directly? AngelCheck wraps Claude with custom-engineered prompts that incorporate pattern recognition from 25 years of venture outcomes, stage-calibrated scoring, and multi-layer quality assurance to catch hallucinations and math errors.
- • Extracts pitch decks and video transcripts (if provided)
- • Scores 8 investment criteria with evidence
- • Identifies red flags and risks
- • Applies VC frameworks (risk tier, power law, exit math)
- • Generates bull/base/bear case scenarios
Multi-Layer Quality Assurance
Three automatic quality checks catch errors before you see the analysis:
- • Accuracy Check: GPT-4o-mini verifies facts, catches hallucinations
- • Auto-Retry: If critical issues found, re-runs with correction hints
- • Text Polish: Final grammar and formatting pass
Get Your Analysis & Ask Questions
Review scores, recommendation (STRONG BUY → STRONG PASS), red flags, VC frameworks, financial analysis, and position sizing recommendations.
AI Chat Assistant: Ask follow-up questions directly on the analysis page. The AI knows the full context of the deal and can clarify scores, explain reasoning, or dive deeper into specific aspects. Available on both individual analysis pages and when comparing multiple deals side-by-side.
What Makes Our AI Different: Pattern Recognition from Real Outcomes
Our AI doesn't just analyze your memo in isolation. It recognizes patterns from 25 years of venture outcomes - studying both unicorn successes and spectacular failures.
What Worked: Unicorn Patterns
Airbnb (2009 Seed) - Would score 7-8/10
Early metrics: <$200/week revenue, rejected by 7 VCs
Signal: Strong founder conviction (lived the problem) + network effects moat
Contrarian insight: "People will pay to sleep in strangers' homes"
Key lesson: Low traction + high conviction + defensible moat = potential unicorn
Stripe (2010 Series A) - Would score 8.5-9/10
Early metrics: Developer-loved product, fast YoY growth
Signal: Collison brothers (technical depth) + perfect founder-market fit
Moat: API ecosystem lock-in, developer experience
Key lesson: Founder-market fit + technical depth + massive TAM = exceptional opportunity
Slack (2014 Series B) - Would score 9/10
Early metrics: $2M ARR, 200% YoY growth, 93% daily active usage
Signal: Stewart Butterfield (prior Flickr exit) + exceptional retention
Moat: Network effects + high switching costs
Key lesson: Product love + retention + network effects = sustainable growth
Uber (2011 Seed/Series A) - Would score 7.5-8.5/10
Early metrics: Rapid city-by-city expansion, strong unit economics
Signal: Supply-side network effects (more drivers = better service)
Risk: Regulatory uncertainty, but massive TAM justified risk
Key lesson: Network effects + large TAM can overcome regulatory risk
Figma (2015 Seed/Series A) - Would score 8/10
Early metrics: Designer love, real-time collaboration
Signal: Technical moat (real-time sync in browser), collaborative editing
Contrarian insight: "Designers will use browser-based tools"
Key lesson: Technical innovation + solving collaboration problem = category creation
What Didn't Work: Failure Patterns
Homejoy (2012-2015) - Would score 6-7/10, Failed
What looked good: $1M+ ARR, rapid growth, YC-backed, large TAM
Red flag missed: No defensibility - cleaning is commoditized
Why it failed: High churn, CAC > LTV, marketplace with no moat
Key lesson: Growth + large TAM ≠ sustainable business without defensibility
Color (2011) - Would score 5-6/10, Failed
What looked good: $41M seed round (record at the time)
Red flag missed: No product-market fit, no users, unclear value prop
Why it failed: Raised too much too early, tried to force adoption
Key lesson: Massive funding without PMF = death by overspending
Quibi (2018-2020) - Would score 6/10, Failed
What looked good: $1.75B raised, Hollywood pedigree (Jeffrey Katzenberg)
Red flag missed: Wrong team for mobile video (Hollywood ≠ tech product)
Why it failed: Ignored market feedback, premium content didn't fit mobile
Key lesson: Wrong team for domain + ignoring users = expensive failure
Yik Yak (2013-2017) - Would score 7/10, Failed
What looked good: Viral growth, college campus penetration
Red flag missed: No business model, no path to monetization
Why it failed: Anonymity enabled toxicity, couldn't monetize
Key lesson: Viral growth without business model = unsustainable
Juicero (2013-2017) - Would score 4-5/10, Failed
What looked good: $120M raised, beautiful hardware design
Red flag missed: Solved a fake problem with over-engineered solution
Why it failed: $400 juicer when you could squeeze packets by hand
Key lesson: Over-engineering a fake problem = expensive mistake
Pattern Synthesis: What Separates Winners from Losers
Winners Had:
- • Network effects OR technical moat OR high switching costs
- • Solved real, painful problems
- • Perfect founder-market fit
- • Path to defensible scale
Losers Lacked:
- • Defensibility (competed on service quality alone)
- • Solved fake problems or ignored users
- • Wrong team for domain
- • No business model or path to profitability
The 8 Investment Criteria We Score
Every deal is evaluated across 8 criteria with evidence-based scores (1-10).
1. Founder Quality
Founder-market fit, credentials, conviction, prior exits, domain expertise
2. Market Size
TAM/SAM/SOM, market growth rate, addressable opportunity
3. Product/Technology
Product vision, technical moat, 10x improvement potential
4. Traction
Revenue, growth rate, customer metrics (stage-calibrated for seed vs Series A)
5. Business Model
Path to profitability, unit economics, monetization strategy
6. Competition
Moat depth (network effects, patents, switching costs), competitive differentiation
7. Timing
Market timing, "why now", tailwinds/headwinds, regulatory landscape
8. Exit Potential
Acquisition targets, exit multiples, liquidity timeline, M&A trends
Scoring Calibration: Scores are calibrated to realistic angel-stage distributions. A 7/10 means "better than 60% of angel deals" (top 40%) - not a "C grade." A 9/10 is exceptional (top 10%).
Privacy & Security
Your deal information is sensitive. Here's how we protect it:
LOCAL Anonymization
Company names, founder names, locations, and financial details are removed using NLP (natural language processing) on your device - not in the cloud. OpenAI never sees the raw sensitive data.
Post-Analysis Leak Scrubbing
After AI analysis completes, we scrub the results again to catch any potential entity leaks. Any remaining company/founder references are replaced with generic placeholders.
No Data Selling
We never sell your data to third parties. Your memos are stored encrypted and only accessible to you.
Educational Purpose Only
AngelCheck provides AI-powered analysis for educational purposes only. This is not financial advice. While we use multi-layer quality assurance (accuracy checking, auto-retry, text polishing) to minimize errors, AI analysis should always be verified independently. Always consult a licensed financial advisor before making investment decisions. Angel investing is inherently risky - 90% of startups fail.