Your AI agent is about to
buy $150,000 of one stock.

You told it the limit was $25,000. Nothing is checking. SYGNL checks every order before it reaches your broker, and stops the ones that break your rules.

api.sygnliq.com/v1/trade/preflight LIVE
Press a button. This is the real engine answering right now.

Four answers. That is the whole product.

ALLOW

The order is fine. Send it.

REDUCE

Too big. Here is the size that fits.

CONFIRM

Risky. A human should look first.

BLOCK

No. This one does not go out.

We pointed a real agent at a real broker

Five orders. Two went through. Three did not. This ran against a live Alpaca account, and you can run it yourself.

The agent asked for
Answer
What happened
Buy $5,000 SPY
ALLOW
Sent. Order 647b9334
Buy $150,000 NVDA
REDUCE
Sent at $25,000. Six times over the limit, cut to size
Buy $10,000 TSLA
BLOCK
Stopped. The strategy failed its audit
Buy $8,000 SPY
BLOCK
Stopped. Already down past the daily loss limit
Buy $20,000 IWM
CONFIRM
Held. Fees would eat most of the profit
what if the agent lies about being approved? CAUGHT
the agent edits its own approval:  $5,000 → $500,000
your code checks the signature…

  signature does not match. order never sent.

Every answer is signed and tied to that one order. Edit one character and the check fails.

Add it in one call

Before your agent sends an order, ask us. No signup to try it.

curl -s https://api.sygnliq.com/v1/trade/preflight/demo \
  -H 'content-type: application/json' \
  -d '{"order":  {"symbol":"NVDA","side":"buy","notional_usd":150000,"order_type":"market"},
       "policy": {"max_notional_usd":25000,"daily_loss_limit_usd":2000}}'
{ "decision": "REDUCE",
  "approved_notional_usd": 25000,
  "reasons": [{"code":"NOTIONAL_CAP","message":"notional capped at policy max 25000.00"}] }
RESTMCP under a second$0.01 a check works with any broker

Full docs and every rule you can set →

Who it is for

Agent builders

You know that telling a model to be careful is not a safety feature. Free while you are on paper trading.

Start here →

Prop firms

Your traders are starting to run agents. Enforce your rules on every order, and keep a signed record of each one.

Talk to us →

Brokers and platforms

Opening up to agents needs a safety story. Be the platform that checks the order, not the one that sends a notification after.

Talk to us →

Why we built it

I built a trading system that won 66% of its trades on paper. Before risking money on it, I rechecked it against real market prices. The 66% was fake. With real fills it won 14%. I threw it away and kept the tool that caught it.

That tool is now this. How it works →

What we are not going to pretend

Nobody pays for this yet. The Alpaca run above is our own example, not a customer. We cannot show you money saved at a real firm, because there is no real firm yet. We would rather say that than round it up, since catching rounded up numbers is the entire job.

Questions

Is the broker not already doing this?

No. Brokers cap how much an agent can lose and text you afterward. Robinhood gives the agent a separate account and sends a notification per trade. Nothing asks whether the order fits your strategy.

Which brokers work?

All of them. We answer, your code enforces. Our example runs on Alpaca. The same pattern works with Robinhood or anything else.

Is this investment advice?

No. We never pick stocks or predict prices. We check the order against rules you set.

How fast is it?

Under a second, and the answer expires after 60 seconds so nothing gets reused.