Swarm Intelligence Tiers
More agents = more perspectives = better verdicts. Connect a Solana wallet to unlock the full adversarial swarm.
| Feature | Free Scout Scan | Tier 1 Investigator |
|---|---|---|
| Access | No wallet needed | Connect Solana wallet |
| Swarm Agents | 2 (Technician + Security) | 6 + Devil's Advocate |
| AI Models | Gemini Flash | GPT-4o Mini, Claude Haiku, Gemini Flash, Gemini Pro, Grok |
| Analysis Categories | Technical + Safety | Technical, Safety, Tokenomics, Social, Macro, DA |
| Debate Protocol | — | ✅ When agents disagree |
| Evidence Depth | 5 findings shown | Full evidence locker |
| On-Chain Storage | — | ✅ Verdict stored as Solana PDA |
| Daily Scans | 3 | 15 |
| Accuracy Estimate | ~70% | ~85% |
| Avg Scan Time | ~10s | ~20s |
| Score Consistency | ±10–15% | ±5–8% |
Technical Deep Dive
Free — Scout Scan
No wallet requiredSwarm Agents (2)
- 📊 TechnicianBot — On-chain metrics, contract maturity, trading patterns
- 🔒 SecurityBot — Smart contract audit, rug pull detection, honeypot scan
Architecture
- 🧠 Model: Gemini Flash (both agents)
- ⚡ Architecture: 2-agent parallel analysis
- 🎯 Output: Score (0-100), Grade, evidence trail
- ⏱️ Speed: ~10 seconds
Limitation: With 2 agents and no debate protocol, Scout Scans can miss nuances that full adversarial consensus catches. Scores may vary ±10–15% between runs.
Tier 1 — Investigator
Connect Solana walletSwarm Agents (6 + DA)
- 📊 TechnicianBot — On-chain metrics & pattern analysis
- 🔒 SecurityBot — Contract audit & vulnerability scan
- 💰 TokenomicsBot — Supply, distribution, inflation analysis
- 🐦 SocialBot — Real-time X/social sentiment via Grok
- 🌍 MacroBot — Market conditions & sector context
- 😈 Devil's Advocate — Adversarial peer review of all agents
AI Architecture
- 🧠 Models: GPT-4o Mini + Claude Haiku + Gemini Flash + Gemini Pro + Grok
- ⚡ Architecture: Parallel multi-agent with cross-referencing
- 🔥 Debate Protocol: Agents flag disagreements → real-time debate
- 🔄 Consensus: DA challenges strongest claim, 1-3 debate rounds
- 🔗 On-Chain: Verdict hash stored as Solana PDA
- 🎯 Accuracy: ~85% (multi-model + DA correction)
- ⏱️ Speed: ~15-25 seconds
Key advantage: 5 different AI models from 4 providers catch different things. The debate protocol catches ~15% of errors that single-model analysis misses. On-chain verdict storage creates a permanent, verifiable record on Solana.
Why More Intelligence Matters
Different models catch different things. Gemini excels at code, Grok at social data, Claude at reasoning.
More perspectives = fewer blind spots. The Devil's Advocate forces every finding to be defended with evidence.
Each additional perspective narrows the error margin. Multi-agent debate outperforms any single model.
Future tiers will scale up to 20+ agents with more advanced frontier models, real-time monitoring, and API access.