Finance is a natural fit for agents: dense information, long decision chains, and a real need for traceability. SolonAI's goal is to combine Solana on-chain asset data with AI agents, giving everyday users asset diagnosis, strategy generation, risk management and automated execution in one place.
Why multi-agent
A single agent juggling on-chain data, allocation and risk tends to spread itself thin. Split the job across single-purpose agents and let an orchestrator coordinate — reliability and explainability both improve.
The agent team
- Diagnosis agent — reads a wallet, analyzes asset distribution, concentration and liquidity.
- Strategy agent — turns diagnosis plus market data into personalized allocation advice.
- Risk agent — stress-tests the advice and sets stop-loss and circuit-breaker rules.
- Execution agent — executes on-chain via contracts once authorized, logging every transaction.
A key design trade-off
The hardest part is shared state. We use one "context snapshot" threaded through every agent: diagnosis writes to it, strategy reads it and appends advice, risk validates the advice — each step leaves a structured, auditable record.
diagnosis ──write──> context snapshot ──read──> strategy
│
risk <──validate──┘ └──> execution
Why Solana
Low fees and fast confirmations make Solana a good fit for the small, frequent on-chain operations of automated execution. And an agent's traceability fills exactly the trust gap that "black-box AI" leaves in finance.
Next steps
We're working on automatic negotiation between agents, and exploring trusted execution environments so the "AI decision + contract execution" chain gets safer and more transparent.