Agents propose.
Arden decides.
A control plane for quant teams to test ideas, govern agent actions, and inspect every paper decision.
PRODUCT / SAVED SERVICE EVIDENCE
See the decision.
A real service result from synthetic data. No broker order was sent.
NOT A LIVE FEED
Cross-sectional momentum v1
2024-01-02 — 2025-12-31 · Arden synthetic daily fixture
RUN FINGERPRINT 0a10695b1c64f8a2cb97bfa3cc4b93966aae200eb84688003d118779a728ac1e
BUY 1,000 shares of AAPL.
Rejected → Safe Envelope → exact replay · 2024-05-13
01 / AGENT PROPOSAL
An agent requests 1,000 shares of AAPL.
Saved synthetic PAPER example. No order reached a broker.
02 / DETERMINISTIC RISK GATE
Arden rejects the 1,000-share intent.
MAX_ORDER_NOTIONALOrder $190,045.30 exceeds $20,000.00 limitBUYING_POWEROrder and fees exceed available cashMAX_POSITIONResulting position 180.5% exceeds 20% limitMAX_EXPOSUREResulting gross exposure exceeds 80% limitPOLICY v1 · PORTFOLIO SNAPSHOT 329624a823df56437927
03 / SAFE ENVELOPE · SAME SNAPSHOT
105 shares of AAPL fit policy.
The next share fails MAX_ORDER_NOTIONAL.
Snapshot preview only; every PAPER order is rechecked before execution.
04 / DECISION REPLAY · SAVED RECEIPT
EXACT MATCH: the saved inputs produce the same rejection.
Inspect complete saved inputs
{
"intent": {
"symbol": "AAPL",
"side": "buy",
"quantity": 1000,
"client_order_id": "seed-risk-proof"
},
"portfolio": {
"cash": "58800.14562884",
"positions": {
"AMZN": {
"quantity": 0,
"average_cost": "0",
"realized_pnl": "79.6821001999999999999999990"
},
"MSFT": {
"quantity": 0,
"average_cost": "0",
"realized_pnl": "1128.421137880000000000000007"
},
"NVDA": {
"quantity": 0,
"average_cost": "0",
"realized_pnl": "428.2684107999999999999999989"
},
"JPM": {
"quantity": 68,
"average_cost": "179.0854130869879518072289157",
"realized_pnl": "477.5282389351807228915662645"
},
"AAPL": {
"quantity": 0,
"average_cost": "0",
"realized_pnl": "-1183.876074259999999999999998"
},
"META": {
"quantity": 37,
"average_cost": "383.983568205",
"realized_pnl": "90.665423585"
},
"GOOGL": {
"quantity": 97,
"average_cost": "163.2509638639175257731958763",
"realized_pnl": "0"
}
},
"fees": "20.428473"
},
"prices": {
"AAPL": "190.0073",
"MSFT": "472.3659",
"NVDA": "48.1916",
"AMZN": "156.0183",
"GOOGL": "161.6081",
"META": "416.7665",
"JPM": "226.9135",
"SPY": "536.5030"
},
"execution_price": "190.04530146",
"fee": "19.004530",
"policy": {
"max_position_pct": "0.20",
"max_gross_exposure": "0.80",
"max_order_notional": "20000",
"max_daily_loss_pct": "0.03",
"max_positions": 8,
"allowed_symbols": [
"AAPL",
"MSFT",
"NVDA",
"AMZN",
"GOOGL",
"META",
"JPM",
"SPY"
],
"denied_symbols": []
},
"timestamp": "2024-05-13T21:00:00Z",
"policy_version": 1,
"running": true,
"killed": false,
"day_start_equity": null
}- Policy
- v1
- Estimated notional
- $190,045.30
- Portfolio snapshot
329624a823df56437927- Violation codes
- MAX_ORDER_NOTIONAL · BUYING_POWER · MAX_POSITION · MAX_EXPOSURE
- Policy
- v1
- Estimated notional
- $190,045.30
- Portfolio snapshot
329624a823df56437927- Violation codes
- MAX_ORDER_NOTIONAL · BUYING_POWER · MAX_POSITION · MAX_EXPOSURE
Receipt rd_95357a8ace112d1c4ff26f55 · arden-risk-v1 · replay uses the stored market, portfolio, and policy. Replay never submits an order.
SYNTHETIC ARTIFACT 97d71480-9029-4468-b1e9-5bbe11111565:ord-00027 · Inspect machine-readable proof ↗
Momentum · research book
Session 95 of 522 · replay close 2024-05-13
MANUAL SIMULATION · No broker connection or worker heartbeat implied by this saved artifact.
API & agent access
Authenticated workspace tokens, HTTP and local stdio MCP.
GET /api/v1/backtests
GET /api/v1/backtests/{id}/verify
POST /api/v1/orders/propose
GET /api/v1/orders/{id}/decision-replay
GET /api/v1/audit-eventsReal routes. Research and decisions stay scoped to the authenticated workspace.
arden_list_strategies()
arden_get_backtest(run_id)
arden_propose_paper_order(...)
arden_replay_paper_decision(order_id)
arden_get_audit_events()MCP tools use the same API. Paper writes require explicit client opt-in.
WHAT WE ARE TESTING NEXT
Independent strategies.
Shared capital.
One policy boundary across quant systems and agents. Multi-strategy capital arbitration is a design-partner hypothesis, not a beta feature.
Discuss the workflow ↗THE SYSTEM / HOW ARDEN WORKS
One decision path.
Every interface.
Whether a request comes from the UI, an agent, or a Python client, it goes through the same authenticated services and risk checks.
The free hosted beta has no continuous worker. Broker reconciliation needs the separate worker; the public demo uses saved synthetic paper evidence.
PROGRAMMABLE BY DESIGN
One API.
One policy.
HTTP and local MCP clients reach the same workspace-scoped services.
Read developer guide ↗POST /api/v1/orders/propose
→ REJECTED · 4 checks
→ safe_envelope.max_quantity: 105
GET /api/v1/orders/{id}/decision-replay
→ EXACT_MATCH · no order submittedBUILT FOR QUANTITATIVE TEAMS