Autonomous AI trading agents on Base, Solana & Robinhood.
Users & job-to-be-done
Traders who want autonomous agents without writing code — and refuse to fund a strategy they cannot simulate first.
Problem
On-chain trading tools were either coder-only or “AI” wrappers with thin risk controls. Capital went live before the strategy was legible.
Insight
The purchase decision is not “open an agent” — it is “trust this agent with funds.” Simulation, archetypes and a multi-layer risk engine have to sit before deposit.
Bet
A no-code platform with Sniper / Quant / Tracker agents, pattern crafting, simulate-before-capital, non-custodial wallets and multi-network execution (Base, Solana, Robinhood).
My role
Solo builder: product definition, UX, frontend and backend end to end. Owned agent archetypes, risk requirements and the simulate → fund → run funnel. AI is both the product surface and a lever in how I specify and iterate.
Journey
- Discover: pick an agent archetype (Sniper / Quant / Tracker)
- Configure: no-code pattern rules (age, mcap, RSI, flow signals…)
- Validate: simulate the agent before risking capital
- Transact & run: fund non-custodial wallet → live execution & PnL
Architecture
- One Go scanner runs live and simulated trading with injected executors, so sim/live parity holds by design.
- Layered risk engine: launch-safety gates, pre-buy re-checks, per-agent loss rules and circuit breakers.
- One chain registry executes on Base, Solana and Robinhood Chain via Jupiter, 0x and Uniswap Universal Router.
Tradeoffs
- Three archetypes over a freeform strategy IDE.
- Simulate-before-capital as a hard gate on the funnel.
- Same agent model across Base, Solana and Robinhood.
Validation
- Simulation as a productized experiment: users can reject a strategy before deposit.
- Password-gated F&F access to learn real agent configuration pain before a public flood.
- Scale signals from execution infrastructure: 358M+ swap volume across 189K+ tokens scored.
Outcome
North star: Agents that users simulate, fund and keep running with risk rules on
A no-code AI trading platform where anyone can build Sniper, Quant or Tracker agents, simulate them, fund non-custodial wallets and run across Base, Solana and Robinhood — backed by a 4-layer risk engine.
Skills & know-how
What this proves: that I can design high-stakes transactional AI products: clear archetypes, pre-purchase validation, multi-network execution and risk controls — not just a chat wrapper on a swap API.