CONCEPT · PHASE 0 ENGINE BUILT

Every agent has a plan.Every plan has an exit.

Build trading agents without code. Prove them in a verifiable Arena. Deploy them only under risk limits you control.

DETERMINISTICSame events in, same decisions out.
REDUCE-ONLY EXITSRisk halts never trap you in a position.
HASH-CHAINEDEvery decision is a tamper-evident record.
NO LEVERAGESpot swaps only at launch.
The loop

Create. Compete. Deploy.

One product loop from an idea to a constrained live agent. Winning a competition never hands an agent control of your capital.

Agent Studio

Turn a trading idea into a structured, versioned strategy. Pick an archetype, define the tradeable universe, entry rules, capital limits and a multi-step exit plan.

  • Momentum, range, breakout and position-guardian templates
  • Allowlist-first universe with liquidity and impact filters
  • Natural-language setup that never changes limits silently
archetype momentum
universe allowlist · 2 mints
entry ret5m ≥ +3.00%
exit +8% → sell 30%
exit +20% → sell 40%
stop −6.00% · time 6h
Under the hood

The policy is the product. Not the bot.

Most trading bots are built around order execution. ExitPlan is built around what an agent is permitted to do, how its decisions are evaluated, and whether its results can be checked.

Agent DNA

Every strategy version has a permanent identity.

Change any field and you get a different hash, so old results stay attached to the old version. Try it: edit the stop-loss.


        
…

Computed in your browser with the same canonical encoding as the Phase 0 engine (engine v0.0.2, DSL 1.0). A hash proves a configuration was not altered. It does not prove market data was accurate.

Risk policy

Limits, made concrete.

Largest single entry
Largest initial exposure
New entries stop at

Thresholds, not guarantees. Prices can gap, swaps can fail, and losses can exceed a daily stop.

Arena

Ranked on risk-aware results.

Drawdown gates disqualify reckless strategies instead of rewarding them. ILLUSTRATIVE · FICTIONAL

01Velocity V3DD 8.1%+14.2%
02GreenlineDD 5.4%+11.8%
03ExitMasterDD 7.0%+9.6%
DQAll-In LarryDD 41.3%+38.0%
Gate drawdown capTies lower drawdown winsVersions frozen at entryIdle zero return is valid
Proof ledger

Decisions you can replay.

Every decision and fill is appended to a hash-chained log. Cash and positions are rebuilt from it, so a tampered record breaks the chain.

#0 DEPOSIT 1,000.00 hash 5ac1…e07
#1 DECISION BUY entry prev 5ac1…e07 · hash 9d42…b1f
#2 FILL buy 100.00 → qty prev 9d42…b1f · hash 17ee…3c0

Illustrative hashes. Replay verification runs in the Phase 0 engine today.

Live view

Watch an agent reason.

Every step an agent takes is a logged decision with a reason, including the ones where it does nothing. This is a scripted simulation of that log.

agent: momentum-hunter · v1.0SIMULATED

    
Roadmap

Built in the open, gated by evidence.

Each phase has a release gate. Live trading only ships after an independent security review, and only if earlier phases show real demand.

BUILT

0 · Engine foundation

Deterministic strategy engine, paper broker, hash-chained ledger and replay tool. Verified on synthetic data.

NEXT

1 · Agent Studio

No-code agent creation, forward paper trading, decision history and equity tracking.

PLANNED

2 · Arena alpha

Seasons, frozen entries, public profiles and reproducible standings.

GATED

3 · Controlled live beta

Segregated wallets and capped deposits after security review and incident drills.

LATER

4 · Strategy network

Publish, fork and verify strategies. Only if users show demand.

What we won't promise

Trust is a feature.

No guaranteed returns.A better process is not a better outcome. Agents can and will lose money.
A stop-loss is not insurance.It may fill at a worse price, or not at all. We show that plainly.
Simulated is not real.Paper results rely on modeled execution. We label them and disclose the limits.
No keys to your main wallet.Live agents use separate, purpose-funded wallets that you can revoke.