About VerdictSwarm
The first multi-model adversarial AI system for crypto token security β built on Solana. We don't ask you to trust one model β we ask you to trust the consensus.
Built on SolanaWhat is VerdictSwarm?
VerdictSwarm is an AI-powered token security platform built on Solana that deploys a swarm of 20+ specialized AI agents to evaluate any token's risk profile. Unlike single-model security tools like GoPlus or RugCheck, VerdictSwarm uses five top-tier AI providers β Gemini, Grok, Kimi, Claude, and ChatGPT β that independently analyze different aspects, then engage in adversarial debate until reaching consensus.
Every verdict is stored immutably on Solana as a PDA (Program Derived Address), creating a permanent, verifiable record of AI security analysis. Where other tools give a single perspective, VerdictSwarm verifies it's not a rug.
The Agent Swarm
Up to 20+ agents across 4 tiers, powered by 5 AI providers
Phase 1 β Scout (Free)
Technician
On-chain metrics, contract maturity, trading patterns, liquidity depth analysis.
π§ Gemini FlashSecurity
Smart contract audit β rug pull patterns, ownership risks, honeypot detection, proxy contracts.
π§ Gemini FlashPhase 2 β Investigator (Tier 1)
Tokenomics
Supply distribution, inflation mechanics, holder concentration, vesting schedules, burn analysis.
π§ Gemini ProSocial Intel
Real-time community sentiment, social signals, narrative trends, influencer activity, bot detection.
π§ Grok (xAI)Macro Analyst
Market conditions, sector trends, correlation analysis, external risk factors, regulatory signals.
π§ Gemini ProDevil's Advocate
Challenges every positive finding. If the swarm is too bullish, this agent pushes back with counter-evidence.
π§ ClaudePhase 3 β Prosecutor & Grand Jury (Tier 2-3)
VisionBot
Visual analysis of website screenshots, social media imagery, and marketing materials for red flags.
π§ Gemini Pro VisionLLM ScamBot
Adversarial pattern recognition trained on 10,000+ known scam signatures, fake team detection, plagiarism analysis.
π§ Kimi + CodexVerdict Synthesizer
Final consensus engine β weighs all agent findings, resolves disagreements, produces the unified verdict score.
π§ Multi-model ensembleOn-Chain Recorder
Stores every verdict immutably on Solana as a PDA. Creates permanent, verifiable audit trail on-chain.
π§ Solana AnchorAdversarial Debate Protocol
When agents disagree on a score by more than 2 points, the swarm triggers a real-time adversarial debate. Agents present evidence, challenge each other's findings, and iteratively update their scores using Bayesian convergence until reaching consensus β or declare a split verdict. This is what makes multi-agent analysis fundamentally more reliable than any single-model tool. Multi-model disagreement IS the signal.
Our Technology: Proprietary Adversarial Consensus Engine
The core of VerdictSwarm is our Proprietary Adversarial Consensus Engine β our internal methodology for orchestrating agent debates, weighting evidence, and converging on a verdict you can trust. It is our secret sauce, continuously refined as frontier models evolve.
We continuously optimize our agent prompts, debate flow, and scoring methodology across Gemini, Grok, Kimi, Claude, and ChatGPT (Codex). As new model releases emerge, the engine is upgraded so the swarm gets sharper over time.
Solana Integration
SPL Token Scanning
Native Solana token analysis via Helius RPC β deep on-chain data, holder analysis, program verification
On-Chain Verdicts
Every scan stored as immutable PDA on Solana via Anchor program β permanent, verifiable, auditable
Tiered Access
Wallet-gated tiers β connect your Solana wallet to unlock deeper multi-agent analysis
Solana Speed
Sub-second finality for on-chain verdict storage β no waiting for block confirmations
Why Trust the Swarm?
5 AI Providers
Gemini, Grok, Kimi, Claude, ChatGPT β no single model dependency
Adversarial Debate
Agents challenge each other when they disagree β no blind consensus
On-Chain Proof
Every verdict stored on Solana β immutable, verifiable, permanent
The Team
Built autonomously by AI agents using multi-model orchestration. Designed for the AI agent era: distributed intelligence, no single point of failure, consensus over individual opinion.
AI Coordinator
@VswarmAi on πAutonomous AI coordinator that orchestrates the swarm, manages operations, and serves as the project's public interface with the community. Builds in public.
The Swarm β 5 AI Providers
Gemini Β· Grok Β· Kimi Β· Claude Β· ChatGPT (Codex)Five top-tier AI providers working in concert: Google Gemini for deep security analysis, xAI Grok for real-time social intelligence, Moonshot Kimi for research and cost-effective analysis, Anthropic Claude for reasoning and architecture, and OpenAI Codex for code analysis and development. These models independently analyze, debate, cross-verify, and reach consensus before rendering verdicts.
Always Evolving
AI moves fast β and so do we. When Google releases a new Gemini, when xAI upgrades Grok, when Anthropic ships a new Claude β we integrate it. Our architecture is model-agnostic: the swarm continuously evolves with the latest and most capable AI models. Your scans automatically get smarter over time.
VerdictSwarm's core technology is proprietary and protected under applicable intellectual property laws.
βTrust the consensus.β