Market-facing agents. Governance-facing rails.

Research, signal monitoring, and execution run by agents inside rails you define: scoped identity, explicit approval boundaries, hard spend limits, and an audit trail on every decision — the same discipline we prove on customer communications, pointed at markets. And it isn't theoretical: customer zero is our own account, trading live.

Discuss trading rails See it running

The problem

Autonomy plus money is exactly where discipline usually dies.

A trading bot with API keys and no governance is a headline waiting to happen. But the answer isn't "no autonomy" — it's the same answer as everywhere else in this platform: identity that's accountable, gates where consequences live, limits enforced by architecture rather than promises, and a ledger you can replay after the fact. Markets don't forgive vibes.

Capabilities

Research on tap

Agents that read everything before you're awake

Filings, feeds, prices, and your own positions — digested continuously into briefs and signals on the data plane. Research agents cite sources and log their reasoning, so a thesis is reviewable, not a hunch.

● the same brief engine runs our exec digests

Signals with thresholds

Watchers that page, not spam

Signal agents monitor the conditions you define and act only within their mandate: notify below the line, propose at the line, and never execute past it without the gate. Every threshold is a policy object, not a code comment.

● threshold-gated watchers run our ops

Gated execution

The approval gate stands between thesis and order

Execution requests hit the same human-in-the-loop machinery that gates our production changes: a decision card with full context, approve or deny, durable ledger either way. Speed where it's safe, a human where it's money.

● the HITL gate is wall-proven daily

Hard limits

Spend caps the agent cannot argue with

Per-agent, per-day, per-position limits enforced by the platform — the same budget machinery that meters our AI spend. Blow-up prevention is a config, not a hope.

● budget caps + degrade chains shipped

What's inside

The manifest.

01 Research agents cited briefs, logged reasoning
02 Signal monitoring policy-defined thresholds
03 Governed execution HITL gates before orders
04 Hard spend limits platform-enforced, per agent
05 Decision audit replayable ledger of every call
06 Scoped identities each agent its own accountable principal
07 Kill switches revoke at the boundary, instant
08 Data-plane analytics performance, attribution, drift

How it lands

Define the mandate

Universe, strategies, thresholds, limits, and who approves what — written as policy the platform enforces.

Paper the loop

Agents run the full research→signal→gate cycle against paper first; you review the decision ledger, not a backtest brochure. Ours earned live the same way.

Go live, gated

Real execution opens with the gates and caps on. Autonomy widens only as the ledger earns it.

Proof

On duty for us, right now.

Customer zero is our own capital. This line trades our own account, live — multi-week continuous runs under hard external risk limits, flat at every close, every session ending in a replayable decision ledger. We publish the discipline, not the returns; the gates and caps it runs behind are the same components proven daily on our public test wall.

— checks passing — failing last full run —

All the receipts Open the live app

Put Trading Agents on your payroll.

Same platform we run our own company on. First line is a purchase — the second is a decision.

Discuss trading rails