Live

Velocis Agent

The AI analyst that watches, tests and explains.

Pattern alerts written in plain language, backtests of the setups your traders actually trade, and what-if answers on their own challenge. Under your brand, with every number computed by a deterministic engine.

velocis agent · sessionengine: deterministic
You:“▌

A replayed session on a sample evaluation account. On a call you get the live product on your markets.

01AI alerts

Alerts used to be a line on a chart. Now they understand a pattern.

A price alert tells you a level was crossed, on every wick, with no idea why it matters. With Velocis Agent a trader describes the setup they actually trade, in their own words, and the engine checks it on every bar.

When it fires, the alert arrives with what happened and how that pattern has played out before.

The old alert

IF NQ > 18,450 THEN notify

A price crossed a line. It fires on every wick, and it knows nothing about context.

The AI alert

“Tell me when NQ sweeps the overnight high, then closes back inside on the 5-minute with volume above its 20-bar average.”

Alert fired · 10:42 ETNQ · 5m

NQ took the overnight high by 6 points, then closed back inside at 18,498 on 1.8× average volume. Since 2021 this pattern has appeared 143 times; price traded back to the overnight midpoint within two hours in 58% of them.

Pattern backtest · ES · 2018 to todaycomputed

› When ES fails an opening-range breakout in the first 30 minutes, what happens over the next hour?

Occurrences
287
Back in range
64%
Median next 60m
−0.24%
Worst case
+1.1%

Next-60-minute return distribution

−1.2%0+1.2%

Definition: break of the 09:30–10:00 high by ≥ 2 ticks, close back inside within 3 bars · 1-minute bars · written out with every result

02Pattern backtests

Test the idea before risking the account.

Traders have hypotheses, not code. They describe a setup in a sentence and get it tested on years of intraday data, with the exact definition the engine used written out next to the result.

No scripting, no spreadsheets, and no number the model made up.

03Challenge intelligence

It knows your rules and the trader's account.

Load your challenge rules, the trader's own trades and your firm's policies, and the agent answers questions no generic chatbot can: why an account failed, which rule is closest to breaking, what would have happened if.

Traders learn from their own data instead of opening a support ticket, and your firm becomes the one that helped them pass.

Challenge rules

50K evaluation · $3,000 target · $2,000 trailing

Trader account

61 trades · 8 sessions

Firm policy

Consistency 40% · news window 2 min

› What if I hadn't traded the first 15 minutes of the day?

You would have passed on day 7, a day before the breach. The 23 opening trades lost $2,140 between them.

Passed · day 7

› Which rule am I closest to breaking this week?

Daily loss. On Tuesday you used 82% of it before 10:15 ET, both times after a loss on the open.

Daily loss · 82%

› If I cut size by a third after two losses in a row, does my drawdown fit?

Yes. The deepest drawdown falls from $2,060 to $1,340, inside the $2,000 limit.

Max DD $1,340

04Market analysis

The research desk in every trader's pocket.

Technical, statistical and fundamental reads computed on live or historical data, explained in plain language.

05Under the hood

The model never does the maths.

Language models are good at understanding a question and explaining an answer, and unreliable at computing one. So Velocis Agent splits the work: the model plans and explains, and everything in between is deterministic code you can test and a log you can read.

  1. 01Trader

    Ask

    A question in their own words, about the market or their own account.

  2. 02Model

    Plan

    Reads the question and picks the tools: data, indicators, rules, backtest.

  3. 03Engine

    Compute

    Typed, tested code runs on Databento, EODHD, Massive or your feed. Every call is logged.

  4. 04Model

    Explain

    Writes the answer from the engine's numbers. It never invents one.

It does not place trades and it does not give advice. It researches, computes, alerts and explains, and it says so when a question is outside what it can compute.

06White-label

Your brand on the front. Our engine behind it.

Your brand
Your logo, colours and domain. Traders see your firm, not ours.
Your data
Your market-data entitlements, your challenge rules and your account data.
Your cloud or ours
Hosted by us, or deployed inside your infrastructure.
Licensed per firm
Priced by seats and data coverage, after a paid pilot shows it works for your traders.
analyst.yourfirm.com

› Backtest 12-month time-series momentum on ES, NQ, ZN, GC, CL and 6E, risk-weighted to a 10% vol target, rebalanced monthly, since 2012.

Futures trend · 6 markets · 10% vol target · 2012 to today

CAGR
9.8%
Sharpe
0.94
Realised vol
10.3%
Max drawdown
−14.2%
Equity+270.2%

Back-adjusted continuous contracts · 1 tick slippage + $2.50/side · computed by the engine

Velocis Agent under a firm’s own brand and domain. Illustrative data.

07How to start

Start with a paid pilot.

Velocis Agent goes live under your brand for a small group of your traders, on your rules and your markets. At the end you get a written report on how they used it, and you decide whether to roll it out.

Discuss a pilot
Duration
Fixed length
Agreed before it starts, with a clear end date.
Price
Fixed price
One fee for the pilot, agreed in writing up front.
Traders
A small group
Chosen by you, on your challenge rules and markets.
At the end
A usage report
What traders asked, what they used, and what it changed.

See it run on your markets and your rules.

Thirty minutes with the live product: your tickers, a sample of your challenge rules, and what a pilot with your traders would look like.

Book a live demo

30 minutes · with the founder · no deck