01 Research fleet
Specialist agents scan structure, flow, volatility, onchain, social and news. Always on.
→ candidates
System overview
One system. Recursive by design.
Every cycle makes the next one sharper.
02 Proof
Backtests on real fees and slippage, out of sample. Most candidates fail here.
→ verified strategies
03 Mandate
Your limits: drawdown, exposure, daily loss, markets. Checked before every order.
→ permitted actions
04 Allocator
Decides where capital goes and how much, across venues and asset classes. Idle is allowed.
→ positions
05 Execution
One path to every venue. Your keys. Agents never hold funds.
→ fills
06 Memory
Fills, misses and regimes are kept. The fleet starts the next cycle from here.
→ the next cycle
Recursion
What recursive means here.
Strategies improve strategies.
Agents fork what passed, vary it, and test again.
Outcomes retrain the fleet.
Live results, not just backtests, shape the next search.
Allocation learns.
Capital follows what keeps proving itself.
Wealth loop
The wealth loop.
Deposit, trade onchain, scale with prop capital, run perps, recycle the profit. New channels open as each is proven.
- Deposit (Planned)
- Onchain (Preview)
- Prop capital (Early access)
- Perpetuals (Live)
- Prediction markets (In development)
- Onchain equities (Planned)
- Real-world assets (Planned)
Agent mesh
A fleet, not a bot.
Specialists share context and challenge each other.
Specialists, not silos
Independent agents, shared state.
Shared memory
Nothing is forgotten between cycles.
Common mandate
Different methods, one objective.
The boundary
What it will never do.
- Trade outside your mandate.
- Hold your funds.
- Deploy anything unproven.
Model agnostic
Any model. Same engine.
Models propose. Everything that proves and executes stays.