Partner and institutional platform

Ship multi-asset trading and AI agentsunder your own brand.

NexusTrade is the execution rail and research engine behind other people's products. You keep the brand, the customer, and the policy layer. We bring brokerage connectivity, backtesting, and agents — with a live-order approval boundary your compliance team controls.

Auditable research · human-gated by default · deploy to your trust boundary

Execution adapters

TradierAlpacaTradeStationPublicPaper — same code path

Asset classes

EquitiesOptionsCrypto
order-rail · partner-sandbox Connected
Proposal → broker, end to end
Every hop recorded and attributable
Traced
  1. 1
    Agent forms a proposal
    Plan, model calls, tool inputs and outputs, cost and latency retained
    NexusTrade
  2. 2
    Your policy engine evaluates
    Position limits, restricted lists, suitability — your rules, your verdict
    Your stack
  3. 3
    A human approves or rejects
    The decision is recorded against the reviewer, never inferred
    Your stack
  4. 4
    The adapter rechecks approval
    The brokerage layer refuses unapproved orders regardless of caller
    NexusTrade
  5. 5
    Routed to the connected brokerage
    Fill, fees and position delta written to the portfolio event ledger
    NexusTrade
Live order approval
strategy · mean-reversion-us · illustrative
Awaiting reviewer
AAPL
Buy · market order
Human gate
Quantity120
Est. notional$25,184
RouteConnected
  • Policy engine — passed, concentration 4.1% against an 8% cap
  • Preflight — buying power and tradability confirmed

Walk-forward v2 · anchored
Train → validate → out-of-sample
Complete
Fold 01TrainValidateOOS
Fold 02Train ↗ValidateOOS
Fold 03Train ↗ValidateOOS
Fold 04Train ↗ValidateOOS
Fold 05Train ↗ValidateOOS
2–8Folds per study
OOS promotion
RustBacktest engine

Optimization runs on training windows only. Candidates are selected on validation. The winner is promoted once to each out-of-sample fold, so the final number is not one the optimizer was allowed to see.

Where the line sits

Your policy layer. Our execution rail.

The hardest part of any partnership is agreeing who owns what. This is the split we run in production, and the one we will hold ourselves to in writing.

Your productYou own it
Brand and UICustomer identityBilling and entitlementsSupport
Your controlsYou own it
Risk and policy engineRestricted listsSuitability rulesSupervision and escalation
The approval gate Agent-originated orders start unapproved. The brokerage adapter rechecks approval itself rather than trusting whatever called it, so a defect in the UI, the API, or an agent cannot release an order. This is the contract between the two halves.
NexusTrade agentsWe own it
Research and strategy authoringBacktestingWalk-forward validationScheduled agents
NexusTrade railWe own it
Order lifecyclePosition and P&L accountingPortfolio event ledger4 brokerage adapters
Yours — we never see or store it Ours — versioned, traced, exportable

What you are actually buying

Three things you would otherwise build.

Ordered by how long each one takes to build yourself. Execution connectivity is first for a reason.

Execution rail

Every broker, one order model.

A single order object routes to any connected brokerage. Auth, reconciliation, partial fills, and position accounting are already solved and already in production.

  • Tradier · Alpaca · TradeStation · Public
  • Equities, Options, Crypto
  • Multi-leg options spreads up to six legs
  • Paper and live share one code path
Research infrastructure

Test the edge out of sample.

Run anchored or rolling walk-forward studies in the Rust engine with a strict train, validation, and out-of-sample sequence.

  • Two to eight fold studies
  • Genetic or systematic optimization
  • One-time OOS promotion
  • Progress, cancel, and checkpoint support
Agents and audit

Reconstruct the decision.

Aurora plans, researches, and authors strategies. Every step is recorded, so when someone asks why an order existed the answer is a record rather than a reconstruction.

  • Owner-scoped agent trajectories
  • Durable portfolio event history
  • JSON and CSV export surfaces
  • Python and TypeScript SDKs, plus MCP

Stated plainly: per-strategy automatic approval is available. When it is enabled no human reviewed that order, and NexusTrade does not imply otherwise. Institutions requiring manual review keep it disabled.

Choose the operating model

Pick the one that matches your trust boundary.

Most partners start whitelabelled and go deeper only when their regulator or their board requires it. The first choice does not lock you in.

WhitelabelMost partnersYour brand, our infrastructurePrivate deploymentOur stack, your infrastructureDirect integrationYour stack, our APIs
Time to first live orderMonthsWeeks
Who sees the customerYou, exclusivelyYou, exclusively
Where strategies liveYoursOur environment, via API
Brokerage relationshipYoursYours
Engineering lift on youHigh — or contract us to build itMedium — you build the UI
CommercialsLicence and supportUsage based
Proven in productionNot yetYes — public SDKs

Integration spec

The answers, before you have to ask.

Written so it can go straight into a diligence memo, including the parts that are not finished. This page is not claiming a compliance certification; it states the technical boundaries a compliance program can evaluate.

Brokerage adapters
Tradier · Alpaca · TradeStation · Public
Asset classes
Equities · Options · Crypto — multi-leg spreads up to six legs, venue dependent
Available upon request
Futures · FX — scoped into the engagement during the boundary review
Not supported
Fixed income — ask about roadmap timing before you scope around it
Execution gate
Agent orders default to unapproved. Enforcement lives in the universal brokerage guard, not the UI.
Approval record
approvedByUser is false → true only after approval, with the decision persisted to the event ledger
Pre-trade risk
Bring your own policy engine, or use ours for position sizing and allocation caps. We do not claim a supervisory system.
Auth model
Scoped partner sessions · brokerage OAuth with server-side token refresh · API keys for programmatic access
Interfaces
nexustrade on PyPI and npm · REST · MCP server · embeddable agent UI
Audit surfaces
Agent trajectories · portfolio event ledger · strategy and order provenance · JSON and CSV export
Validation
Walk-forward v2 — ANCHORED or ROLLING, foldCount 2–8, GENETIC or SYSTEMATIC_SWEEP, with progress · checkpoint · cancel · persisted results
Entity
StarksTech LLC · Georgia, USA
Commercials
Revenue share on whitelabel, licence plus support on private deployment, usage based on direct integration. Exact terms come out of the architecture review.

Engagement model

A scoped pilot with a defined exit.

Four phases, each ending in something you can hold. If phase two fails you have spent weeks and learned the answer, which is the point of running it second.

Phase 01

Boundary review

Map identity, data ownership, brokerage relationships, and who signs off on a live order.

You leave withA written split of responsibility and the integration surface.

Phase 02

Riskiest boundary first

Prove the single hardest thing — usually identity federation or the brokerage handoff — against one narrow workflow.

You leave withA working path, or a clear no before anyone spends real money.

Phase 03

Paper, then live

The same code path with the approval gate on and real reviewers. Scoped sessions and telemetry in place.

You leave withOrders flowing end to end under your brand.

Phase 04

Evidence review

Walk your risk and compliance team through real agent traces, real approvals, and real portfolio events.

You leave withA go or no-go with evidence attached, and commercial terms.

Diligence

What gets asked on the second call.

Answering these here shortens your internal review, which is the only thing that actually moves a deal forward.

Can we keep our own risk and policy engine?

Yes, and it is the shape we prefer. Your engine evaluates before the approval gate; we enforce the gate and route the order. We do not require you to adopt our sizing or allocation controls, and we never overwrite your verdict.

What stops an agent from placing an order on its own?

Agent-originated orders are created in an unapproved state. The check that releases an order to a brokerage lives inside the brokerage layer itself, so it runs no matter which caller initiated the request — the API, the UI, an agent, or a scheduled job.

Per-strategy automatic approval can be enabled deliberately. We will not describe an auto-approved order as human-reviewed, and we will ask you not to either.

Do you support futures and FX?

Upon request. Equities, options, and crypto ship today. Futures and FX are available when scoped into the engagement — raise them on the first call so adapter work lands in the right phase rather than as a late surprise.

Who owns the strategies our customers build?

Your customers, and by extension you. On the private deployment path strategies never leave your infrastructure. On the whitelabel path they live in our environment under partner-scoped access and are exportable in full at any time.

What does it cost?

Whitelabel is a revenue share on the integrated offering rather than a licence fee, so our incentive is that your product works. Private deployment is licence plus support. Direct integration is usage based. Exact numbers come out of the boundary review, once the shape is known.

Are you exclusive to one partner per market?

No. Existing agreements are non-exclusive. If exclusivity in a defined segment matters to you, that is a commercial conversation rather than a technical one.

Next step

Thirty minutes, engineer to engineer.

Tell us where NexusTrade needs to run, what it needs to connect to, and who must approve live execution. We will walk the approval gate, the adapter layer, and the event ledger with your architect.

Email Austin

No sales sequence. If it is not a fit we will say so on the call.

NexusTrade provides technical controls that may support an institution's compliance program. It does not represent regulatory approval, compliance certification, legal advice, or investment advice. Institutions remain responsible for configuring and supervising their own policies, permissions, and execution workflows.