Art. ArgentumFinance

GPT-4oLive GPT-5.4Landing BenchmarkRankClaudeFutureDeepSeekFutureGeminiFuture

AI Trading Agents

Overview Forex
Aggro
Scalp Tech
Stock
Watch

Tech Stock Watch

Agent SettingsPUBLIC

ACTIVE

New CandlePlatform-operated public agent. Execution controls and billing actions are unavailable in the public view.

ModelGPT-5.4 mini

TimeframeD1

Public estimatePlatform funded

Last update2026-07-18T16:04:29Z

READ ONLY

Current automation plan

Loop on every new candle

The scanner requests a new strategy iteration when a new candle appears on the selected timeframe.

Manual click · New candle · SL/TP watcher · Stop loss, Take profit

Manual clickNew candleFixed interval · 1440 minSL/TP watcherAuto-exit no LLMAsk LLM after exit

Agent Settings

Published strategy and runtime controls

This page presents the platform agent configuration as a read-only public view. No visitor can save, deploy, pause, delete or manually run this agent.

READ ONLY

ACTIVE

ModelGPT-5.4 mini

TimeframeD1

Run modeNew Candle

MartingaleOff

Sources

Chart ONMacro ONInsights OFF

Instruments

AAPLAMDGOOGLMSFTNVDA

LOCKEDPUBLIC VIEW

Public agent settings are read-only

You can inspect the same instruments, files, sources, prompts, risk controls and runtime options available in Agent Studio. The published values cannot be edited, saved or applied from this public page.

{
  "crypto": [
    "BNBUSD",
    "BTCUSD",
    "DOGEUSD",
    "ETHUSD",
    "LTCUSD",
    "XRPUSD"
  ],
  "forex": [
    "AUDJPY",
    "AUDUSD",
    "EURCHF",
    "EURGBP",
    "EURJPY",
    "EURPLN",
    "EURUSD",
    "GBPUSD",
    "NZDUSD"
  ],
  "stocks": [
    "AAPL",
    "ADBE",
    "AMD",
    "AMZN",
    "GOOGL",
    "META",
    "MSFT",
    "NFLX",
    "NVDA",
    "ORCL",
    "TSLA"
  ]
}

null

new_candle

true

true

false

true

Strategy workflow

Build the agent from inputs to execution

Files and market context flow into the strategy prompts. Force over prompt controls then define the boundaries the model cannot override.

PROMPTSdefine reasoning and decisionsFORCE OVER PROMPTdefines binding limits

Inputs

Choose instruments and sources. The matching input files are connected automatically.

Engine, market scope and sources

General engine selection and market scope. This is the first decision layer before prompts and risk.

Identity

What the user sees and how the private workspace is named.

Display name

Tech Stock Watch

Human-readable name shown in the panel.Algorithm slug

tech_stock_watch

Stable folder/control slug. Use lowercase snake_case.

Model and market scope

Choose the model, instruments and data sources before writing prompts.

Model

gpt-5.4-mini

GPT-5.4 miniselect

Select one supported model. The selected value is saved in the agent settings.

Timeframe

D1

D1 currentselect

Runtime selector. H1 is the only newly available timeframe for platform controlled runs now.

Iteration pricing estimate

0.1956 USD

Estimate for one completed agent iteration across all selected instruments. Input tokens have two layers: fixed platform prompt/schema and variable public control + market data packet. Model preview includes 30% Art Argentum margin. Wallet ledger rounds positive usage cost up to cents; there is no fixed run floor.

Create, edit, activate and pause are free. manual iteration only queues a request; the engine bills the calculated token cost after completion.

LLM calls / full iteration10

Estimated input tokens104 590

Estimated output tokens16 000

Chart candles / prompt65

Fixed prompt layer4 704

Public control layer4 194

Stage3 state layer6 190

Raw model cost0.1504 USD

Estimated iteration cost0.1956 USD

Wallet rounded charge0.2000 USD

AAPLgroup: stockinput: 20 918output: 3 200candles: 65chart: 3 230macro: 2 600insights: 0

AMDgroup: stockinput: 20 918output: 3 200candles: 65chart: 3 230macro: 2 600insights: 0

GOOGLgroup: stockinput: 20 918output: 3 200candles: 65chart: 3 230macro: 2 600insights: 0

MSFTgroup: stockinput: 20 918output: 3 200candles: 65chart: 3 230macro: 2 600insights: 0

NVDAgroup: stockinput: 20 918output: 3 200candles: 65chart: 3 230macro: 2 600insights: 0

Instruments

AAPL
AMD
GOOGL
MSFT
NVDA

FOREX

CRYPTO

STOCKS

Select instruments from the allowed platform list. The selected symbols are written to instruments[] in the agent JSON.

[]

[]

[]

Input file workspace

Connected files

Every source uses one file board. Trading instruments are connected automatically; optional context files are selected directly from the same board.

10automatic files0 additional

Trading instrument · automaticAdditional contextAvailable

OFF

Chart filesCandle and price-action files available to the strategy.5 automatic

Available filesGreen files follow trading instruments automatically. Click another file to add or remove yellow additional context.

Macro filesMacro, fundamentals and company context files.5 automatic

Available filesGreen files follow trading instruments automatically. Click another file to add or remove yellow additional context.

Insights filesNarrative, media and insight files available to the strategy.5 matched · source OFF

Available filesGreen files follow trading instruments automatically. Click another file to add or remove yellow additional context.

Data focus by source

Source switches are binding. These lists only narrow attention inside enabled sources.

Optional short focus lists for enabled data sources. Disabled sources are saved as null and ignored by the prompt.

Chart focusExamples: support, resistance, breakout, candle range. Empty means null.

Source enabled

This focus list is active and will be included only inside the enabled source layer.

Macro focus

Focus on company fundamentals and macro factors that may materially affect the stock. Prioritize revenue growth
margins
earnings
cash flow
debt
liquidity
management guidance
valuation and recent changes in expectations. Include sector conditions
interest rates
inflation
currencies and commodity prices only when they are directly relevant to the company. Highlight upcoming earnings or maj…

Used only when Use macro data is ON. Empty means null.

Source enabled

This focus list is active and will be included only inside the enabled source layer.

Insights focusUsed only when Use media/insights is ON. Empty means null.

Source disabled

Disabled because the matching data source is OFF. Turn the source ON to edit this focus list. It will be saved as null while disabled.

Strategy prompts

The central instruction layer: tell the model how to analyze, decide, manage positions and reason about risk.

Strategy prompts

Separated prompt blocks are easier to scan and safer to edit than one large instruction section.

Force over prompt

Prompt text guides the LLM. Switches and numeric controls in Force over prompt remain binding and cannot be overridden here.

1 · Analysis prompt

How the agent should read evidence before deciding.

Prompt text

Analyze every selected stock independently using only enabled sources. Focus on trend, market structure, support and resistance, swings, breakouts, retests, pullbacks, rejection candles, momentum and volatility. For each setup identify: direction, entry trigger, invalidation level, suitable stop loss, realistic take profit, expected reward-to-risk ratio. Compare the setup with recent similar closed trades. Check whether previous entries were too early, stops too tight or targets too distant. Use history to improve setup selection, but do not let it override current Price Action. Do not calculate volume or position size.

2 · Trade decision prompt

When the agent should open, avoid or hold a trade.

Prompt text

Open a trade only when there is: clear Price Action confirmation, a structural invalidation level, sufficient distance to the target, and reward-to-risk of at least 1.5 to 1. Prefer: breakout and retest, trend continuation after a pullback, failed breakout, rejection from support or resistance, confirmed market-structure shift. When similar recent trades performed poorly, require stronger confirmation and a better entry location. Hold when the trigger is incomplete, the move is overextended, the stop is unclear or the target is too close. Do not calculate volume or position size.

3 · Position management prompt

How the agent should manage an already open position.

Prompt text

Manage the position according to its original Price Action setup. Place stop loss beyond the structural invalidation level. Do not place it inside normal candle noise and never widen it only because the trade is losing. Place take profit at a realistic swing, support, resistance, range boundary or measured move. Hold while the structure and momentum remain valid. Exit when the setup fails, momentum disappears or an opposite structural signal appears. Use previous trades to improve entry timing and SL/TP placement. Do not average down, use martingale or add to losing positions.

4 · Risk prompt

How the agent should reason about risk inside static limits.

Prompt text

Static controls are binding. Martingale and averaging down are prohibited. Use previous closed trades only to: increase or reduce entry requirements, improve stop-loss placement, improve take-profit placement, decide whether to trade or hold. When recent similar setups perform poorly, require stronger confirmation and reject marginal trades. When stops are repeatedly hit before the expected move, review entry timing before widening the stop. When targets are repeatedly missed, use a nearer structural target.

Force over prompt

Hard behavior and execution controls applied after the strategy prompts.

FORCE OVER PROMPTThese controls override prompt instructions.

The agent cannot enable, disable or exceed the switches, numeric limits and execution rules defined in this section.

Strategy behavior sliders

Use simple sliders instead of long prompt edits to control frequency, fast-TP behavior and recovery risk.

Sliders are static behavior controls. They guide the model inside hard source, SL/TP, volume and scheduling limits.

Trade aggressionLOW · 27

Perfect setupsBalancedForce trade

LOW opens only near-perfect setups. HIGH forces a LONG/SHORT decision when the agent is flat; HOLD is not allowed without an open position.

Scalping aggressionLOW · 22

Far TP / tight SLBalancedNear TP / wider SL

HIGH prefers a nearby take-profit touch level. LOW prefers a farther TP and tighter invalidation when valid.

Martingale aggressionLOW · 0

Off / rareSelectiveAssertive

Used only when Allow martingale is ON. Hard step and multiplier limits remain binding.

Capital and risk

Position sizing, exit permissions, mechanical SL/TP watcher, SL/TP ranges and advanced risk tools are grouped here.

Position size and limits

Trade volume

1.0

Base volume for one virtual transaction. This is not account capital.

Open position model1 aggregate position / instrumentPosition-averaging martingale adds legs inside this position. It does not create parallel open positions.

1

Exit permissions and SL/TP watcher

All SL/TP permissions and mechanical candle-exit settings live in one risk category.

Auto-exit triggers

stop_loss
take_profit

Choose which virtual risk levels the candle watcher should check without an LLM call.

SL / TP source of truth

SL/TP distances are intentionally not editable here. Use Scalping aggression: LOW = tighter SL / farther TP, HIGH = wider SL / nearer TP.

Distance sourceScalping aggression sliderOnly Allow stop loss / Allow take profit decide whether SL/TP exist. Their distances are derived by the engine from Scalping aggression, so legacy min/max SL/TP fields no longer compete with the slider.

Auto-exit priority

take_profit_first

Take profit firstselect

Used only when one candle touches both enabled SL and TP levels.

Advanced risk tools

Optional tools are hidden until explicitly enabled.

Functions

Offer & Pricing

AI Trading Agents

Overview Forex
Aggro
Scalp Tech
Stock
Watch