A verified multi-agent intelligence network built on Ambient Network. Submit a Solana wallet address, token contract, or transaction signature - five specialized AI agents coordinate in sequence to analyze it, each building on the previous agent's findings, and every decision generates a cryptographic SHA-256 proof receipt.
AmbientMind lets you analyze Solana addresses for security threats using a coordinated network of five AI agents powered by Ambient Network inference. Each agent has a specialized role and passes its findings to the next agent in the chain. Every agent decision is cryptographically hashed and stored as a verified proof receipt.
- Orchestrator - Plans the analysis task and coordinates all agents
- Scout - Performs initial pattern analysis and flags risk indicators. Uses tool calling to validate address format and type
- Analyst - Calculates risk score and severity using Scout findings
- Threat - Cross-references known Solana exploit signatures using Scout and Analyst outputs
- Reporter - Synthesizes all findings into a final threat verdict with recommended action
- Connect your Phantom wallet
- Select task type - Analyze Wallet, Scan Token, or Investigate Transaction
- Paste a Solana address or transaction signature
- Select model - Default routing or a specific model
- Watch all 5 agents coordinate in the live console, each receiving the previous agent output
- Every agent decision generates a SHA-256 proof receipt stored in your session
- View all proofs in Proof Receipts, full mission history in Mission History, agent activity in Agent Network
True sequential dependency chain. Each agent receives the previous agent actual output as input - not just the original prompt repeated. If an agent fails, dependent agents auto-skip with a clear failure message rather than producing results without proper input.
Shared state via React Context persists proof receipts and mission history across all pages for the full session.
The Scout agent uses Ambient tool calling to validate address format and type before analysis. Scout calls check_address_format with the submitted address - the tool validates length and base58 character set and returns whether it looks like a wallet, token contract, or transaction signature. Scout incorporates this into its analysis. This demonstrates Ambient API tool use capability alongside standard inference.
- Analyze Wallet - Security pattern analysis for Solana wallet addresses
- Scan Token - Rug pull and contract risk analysis for token contracts
- Investigate Transaction - Security analysis for Solana transaction signatures
Each task type has specialized system prompts per agent tuned to the specific analysis needed.
Supports Ambient default routing (recommended), GLM-5.1 FP8, and Kimi K2.7 Code. Default routing sends no model field and lets Ambient automatically select the most available compute path. Proof receipts record which model was used for each agent decision.
- React + Vite + Tailwind CSS
- Ambient Network API with OpenAI-compatible endpoint
- Phantom wallet adapter for Solana wallet connection
- SHA-256 proof receipts via Web Crypto API
- localStorage for session persistence of proofs and mission history
Every AI decision in AmbientMind is verified through Ambient Network inference. The app uses Ambient's OpenAI-compatible chat completions endpoint with tool calling support.
Clone the repo, add your Ambient API key to .env as VITE_AMBIENT_API_KEY, then run npm install and npm run dev.
Live demo - ambientmind.vercel.app