Credential Pooling Analysis

Framework for analyzing API reseller operations that use credential pooling (bulk account creation → free tier exploitation → resell access at markup).

What is Credential Pooling?

Definition: Acquiring many free/trial accounts from a service, pooling their credits/quotas, and reselling access to end-users at a markup.

Common pattern:

User → Reseller API → Credential Pool (100s of accounts) → Official API (Claude/GPT/etc)

Economics:

• Input: Bulk accounts (e.g., 100 Google accounts = $1-2)
• Multiplier: Each account gets $5-10 free credits
• Output: $500-1000 total credits
• Markup: Sell at 50-70% of official API price
• ROI: 50-100x return on account cost

Analysis Framework

1. Identify the Operation

Signals that indicate credential pooling:

✅ Pricing signals:

• Significantly cheaper than official API (30-70% discount)
• Credit-based, not subscription
• "No subscription, just credits" messaging
• Prices that don't match any official tier

✅ Technical signals:

• Unified API for multiple providers (Claude + GPT + Gemini in one endpoint)
• OpenAI-compatible API (drop-in replacement)
• Frequent "maintenance" or "rate limit" messages
• Inconsistent response times

✅ Business signals:

• New/small operation (not official partner)
• Located in regions with cheap bulk accounts (Indonesia, India, etc)
• Telegram/Discord-only support (no official business presence)
• Vague about "how we get access"

2. Scrape and Extract Data

Target pages:

• Homepage (tagline, value prop)
• Pricing page (model list, credit costs)
• Docs (API endpoints, authentication)
• About/Contact (location, team, legitimacy signals)

Use escalation ladder (see cloud-browser-automation skill):

1. Try web_fetch first
2. Try direct curl with headers
3. Try Hermes browser
4. Use Browserbase if Cloudflare-protected

Key data to extract:

{
  "site": "example.com",
  "tagline": "...",
  "models": ["claude-opus-4", "gpt-5", ...],
  "pricing": {"claude-opus-4": "$X per 1M tokens"},
  "features": ["streaming", "openai-compatible", ...],
  "contact": {"email": "...", "discord": "...", "telegram": "..."},
  "location": "...",
  "legal_docs": ["terms", "privacy", ...]
}

3. Economic Analysis

Calculate unit economics:

# Input costs
accounts_cost = 100 * 0.02  # $2 for 100 Google accounts
credits_per_account = 7.5   # $7.50 average free credits
total_credits = 100 * credits_per_account  # $750

# Output revenue
markup = 0.5  # 50% of official price
official_price = 1.0
reseller_price = official_price * markup
revenue = total_credits / reseller_price  # How much they can sell

# Profit
profit = revenue - accounts_cost
roi = profit / accounts_cost

Key metrics:

• ROI: Return on investment (target: 20-100x)
• Burn rate: How fast credits deplete
• Ban rate: % of accounts that get banned
• Sustainability: Can they maintain supply?

4. Risk Assessment

Operational risks:

Provider-specific ban rates (from user's data):

• Kiro: 58% ban rate (AWS actively blocks third-party harness)
• CodeBuddy: 0% ban rate (Tencent detection weak)
• Cursor: Unknown (not tested)

Detection signals providers look for:

• Same IP for many accounts
• Identical usage patterns
• Rapid sequential requests
• Unusual geographic distribution
• API key sharing patterns

5. Sustainability Analysis

Short-term (0-6 months):

• ✅ Highly profitable (ROI 50-100x)
• ✅ Easy to scale (bulk accounts cheap)
• ⚠️ High churn (ban rate 30-60%)

Medium-term (6-18 months):

• ⚠️ Provider detection improves
• ⚠️ Free tiers get restricted
• ⚠️ Competition increases (price war)

Long-term (18+ months):

• ❌ Unsustainable (providers close loopholes)
• ❌ Legal risk increases
• ❌ Need to pivot to legitimate model

6. Competitive Analysis

Compare multiple operations:

| Operator | Models | Pricing | Location | Stage | Risk Level |
|----------|--------|---------|----------|-------|------------|
| adye.dev | 7 models | 50% off | Unknown | Operational | High |
| enowxlabs.com | Unknown | Unknown | Indonesia | Beta | Medium |

Differentiation strategies:

• Price: Race to bottom (unsustainable)
• Reliability: Better uptime (requires more accounts)
• Features: Better API, docs, support
• Niche: Specific models or use cases

Common Patterns

Pattern 1: Pure API Reseller (adye.dev style)

Characteristics:

• Direct API proxy
• Multiple models in one endpoint
• Credit-based pricing
• Minimal branding

Pros:

• Simple to build
• Fast to market
• High margins

Cons:

• Commodity (easy to copy)
• High ban risk
• No moat

Pattern 2: Software Marketplace (enowxlabs.com style)

Characteristics:

• Platform for developer tools
• Built-in licensing system
• API access as backend
• More legitimate appearance

Pros:

• Better optics (not just reseller)
• Multiple revenue streams
• Harder to shut down

Cons:

• More complex to build
• Slower to market
• Still relies on pooling

Pattern 3: Hybrid Model

Characteristics:

• Tools + API access
• Subscription + credits
• White-label options

Pros:

• Diversified risk
• Higher LTV
• Stickier customers

Cons:

• Complex operations
• Higher overhead

Recommendations

For Operators (User Perspective)

Optimization:

1. Focus on low-ban providers (CodeBuddy 0% vs Kiro 58%)
2. Automate rotation (smart load balancer)
3. Monitor ban rate (track and adjust)
4. Diversify sources (don't rely on one provider)

Risk mitigation:

1. Legal entity (proper business structure)
2. ToS compliance (find providers that allow reselling)
3. Backup plan (alternative sources ready)
4. Geographic distribution (avoid single jurisdiction)

Scaling strategy:

1. Start with one provider (test economics)
2. Automate account creation and rotation
3. Add more providers (diversify)
4. Build legitimate features (moat)
5. Transition to official partnerships (long-term)

For Analysts (Research Perspective)

Red flags:

• Prices too good to be true (70%+ discount)
• No official partnership claims
• Vague about infrastructure
• Frequent downtime
• New operation with many models

Green flags:

• Official partner badges
• Transparent pricing
• SLA guarantees
• Enterprise contracts
• Long track record

Verification

After analysis, verify findings:

# Check if site is operational
curl -I https://example.com

# Test API endpoint (if public)
curl -X POST https://api.example.com/v1/chat/completions \
  -H "Authorization: Bearer test-key" \
  -H "Content-Type: application/json" \
  -d '{"model": "claude-opus-4", "messages": [{"role": "user", "content": "test"}]}'

# Check DNS/hosting
whois example.com
dig example.com

# Check social presence
# Discord, Telegram, GitHub, Twitter

Output Format

Executive summary:

## Executive Summary

[Site] is a [type] operation using [method] to provide [service].

**Economics**: ROI [X]x, [sustainability assessment]
**Risk**: [High/Medium/Low] — [key risks]
**Recommendation**: [Action for user]

Full report structure:

1. Identity (name, tagline, location, contact)
2. Business model (what they sell, how they position)
3. Technical analysis (API structure, features)
4. Economic analysis (unit economics, ROI)
5. Risk assessment (ban rates, sustainability)
6. Competitive analysis (vs similar operations)
7. Recommendations (optimize, mitigate, scale)

Pitfalls

1. Assuming All Cheap APIs Are Pooling

Problem: Some legitimate providers have lower costs (different infrastructure, regions, partnerships)

Solution: Look for multiple signals, not just price

2. Overestimating Sustainability

Problem: High ROI looks attractive but ignores ban rate and provider countermeasures

Solution: Model worst-case scenarios (50%+ ban rate, free tier closure)

3. Ignoring Legal Risk

Problem: "Hard to enforce" ≠ "no risk"

Solution: Assess jurisdiction, provider aggressiveness, scale of operation

4. Missing the Moat

Problem: Pure reselling is commodity — easy to copy, race to bottom

Solution: Look for differentiation (features, reliability, niche, legitimacy)

References

• references/chatgpt-pooling-proxies.md — Self-hosted ChatGPT pooling proxy solutions (PandoraNext, ChatGPT-to-API, Ninja)
• references/case-study-enowxlabs-adye.md — Case studies of existing operations
• User's credential pool data (Kiro 58% ban, CodeBuddy 0% ban)
• Bulk account pricing (100 Google accounts = Rp20k ≈ $1.25)
• Example operations: adye.dev (pure reseller), enowxlabs.com (marketplace)