NOCTURNER
Try dApp

PRE-TRADE INTELLIGENCE FOR TOKENIZED EQUITIES

The market sleeps.
Your risk engine shouldn't.

Nocturner scores the risk of a tokenized stock trade before it executes—combining market state, oracle freshness, liquidity, price divergence, and unpriced news.

Request early access Explore the engine
PRE-TRADE / LIVE SIMULATION ENGINE ONLINE
ASSETNVDAx / USDC
SESSIONCLOSED 06H 12M
RISK SCORE68/100
VERDICTSTALE

Reference price can no longer defend execution.

ORACLE AGE340s
POOL DEPTH0.4× order
LAST CLOSE GAP+2.84%

CATALYST DETECTED / 03:14 UTC New 8-K classified as material. Not yet reflected in oracle mid.

01 / THE GAP

AI/ML
AI CATALYST LAYER
SCORING
RISK SCORING ENGINE
POLICY
ERC-4337 POLICY MODULE
INFRA
ROBINHOOD CHAIN

Tokenized stocks trade 24/7.
The exchanges that price them do not.

When the closing bell rings, onchain markets keep moving. Reference prices age. Spreads widen. Depth disappears. Breaking news changes fair value—with no circuit breaker and nobody watching.

02 / FIVE SIGNALS. ONE DECISION.

Risk, assembled
in real time.

Nocturner turns fragmented market conditions into one inspectable score and a decisive execution verdict.

01

Session state

Knows whether the underlying exchange is open, closed, in pre-market, or approaching a scheduled event.

02

Oracle freshness

Measures how old the reference price is and escalates when the market is moving faster than the feed.

03

Executable depth

Compares your order to actual pool liquidity—not a headline TVL number that disappears on execution.

04

Price divergence

Tracks the distance between the token, oracle mid, and the underlying equity's last defensible close.

05

Unpriced catalysts

Reads filings, halt notices, corporate actions, and news that may not be reflected onchain yet.

THE VERDICT SYSTEM

Readable in a second.
Defensible under scrutiny.

0–24

SAFE

Conditions are inside policy limits.

25–49

WIDE

Execution cost or divergence is elevated.

50–74

STALE

Reference data can no longer defend the price.

75–100

HALT

Policy rejects execution until conditions recover.

Thresholds shown are illustrative while calibration is in progress. Production policy is configurable per user, agent, and protocol.

03 / AI WHERE IT MATTERS

The market has context.
Static rules don't.

AI finds and structures the information a threshold cannot see. Every execution decision remains deterministic, auditable, and reproducible.

Our design principles
01

Catalyst Sentinel

Monitors filings, earnings wires, halt notices, and material overnight news before stale liquidity can react.

02

Corporate Action Extractor

Turns splits, dividends, mergers, ex-dates, and delistings into a structured, continuously maintained calendar.

03

Policy Compiler

Translates plain-English limits into deterministic, auditable guard rules. The model drafts; the contract decides.

04

Anomaly Detection

Learns normal spread, depth, and volume signatures by session, then surfaces deviations static thresholds miss.

05

Risk Explainer

Turns raw factors into a clear reason a person—or another agent—can inspect before acting.

06

Threshold Tuning

Backtests weights against realized outcomes so risk policy can improve as market structure changes.

04 / ONE ENGINE, FOUR SURFACES

Protection wherever
execution begins.

01

Public dashboard

Clear scores, factors, and overnight context for retail traders.

For humans
02

Telegram alerts

Immediate risk changes and overnight gap reports for active traders.

See delivery
03

MCP server

A native risk query and dry-run endpoint for autonomous agents.

For agents
04

ERC-4337 guard

Trades that breach policy revert at the session-key level.

See enforcement

05 / ARCHITECTURE

AI observes.
Policy decides.
The guard enforces.

The separation is deliberate. Probabilistic systems discover context and compile intent. Deterministic infrastructure owns the final decision around capital.

OBSERVEMarket + catalyst signalscontinuous
DECIDERisk score + policyinspectable
ENFORCEERC-4337 guarddeterministic

06 / BUILDING IN THE OPEN

NOW

Read-only scorecard

Collecting historical conditions and exposing transparent factors before enforcement.

NEXT

Agent dry runs

Policy simulation and MCP-native verdicts for agent developers.

LATER

Guarded execution

Session-key enforcement after thresholds are proven against real outcomes.

View the full roadmap

EARLY ACCESS / 2026

Know the risk before the trade.

We're opening the read-only scorecard first while the engine collects the history needed for responsible calibration.

Request access Read the research