FluxGeko
Start 30-day trial
AI trading infrastructure

Win the fill.

Autonomous AI agents on a microsecond execution engine. Hard caps the model can't override. Your keys, your control.

Non-custodial. Your exchange keys never leave your machine's session.

FluxGeko · brand film
Non-custodial by design
Journal-only by default
Caps enforced in-path
Kraken · Binance · Coinbase
01 · The path

Every order takes the same path.

The caps sit in the execution path, not in a settings page.

01

Signal

The agent reads the market and proposes a trade, reasoning attached.

02

Rule-check

The engine checks every cap before anything moves.

In-path
03

Route

Orders route to your exchange over your own keys.

04

Fill

Fills are confirmed, journaled, and attributed.

02 · The cage

The model proposes. The engine disposes.

Hard rules the AI cannot override.

Example values, set by you. Caps are enforced in the execution path. The agent cannot change them.

Max notional

250.00 EUR

Caps the size of any single order the agent can place.

Max open executions

6 concurrent

Limits how many executions can run at the same time.

Daily-loss kill switch

150.00 EUR

Halts dispatch for the day the moment your limit is hit.

Journal-only mode

ON by default

Every decision is logged, nothing is dispatched, until you flip it.

03 · The engine

This isn't a bot. It's an engine.

One dark surface from this page to the product. What you see here is what you sign into.

FluxGeko WS live H
BTC/EUR Live 91,280.10 +124.50
09:41:12 UTC
Agent activity Journal
09:41:02 Agent Proposed SV entry on BTC/EUR
09:41:02 Engine Rule-check passed: caps OK
09:41:03 Engine Routed limit buy to Kraken
09:41:05 Kraken Fill confirmed: 0.0041 BTC
Trading caps
Max notional250.00 EUR
Open execs3 / 6
Daily-loss limit150.00 EUR
DispatchJournal
Simulated data BTC/EUR · Kraken
04 · The agent

Proof, not promises.

Every decision is logged, attributed, and timestamped, before and after it becomes a trade.

Agent decision feed Journal Simulated data
09:41:12 Agent Journaled SV entry on BTC/EUR: caps OK, journal-only mode. BTC/EURSV
Journal entry Trend up on the 1-hour chart, pullback to the 25-bar mean on the 5-minute. Spread and fees leave executable edge, not theory. Caps check: notional 240.00 of 250.00 EUR, open executions 3 of 6. Action: journaled, no dispatch.
09:38:40 Agent Skipped ARB route on ETH/EUR: edge below fee floor. ETH/EURARB
09:35:07 Engine Blocked EMA_PB order: would exceed max open executions. BTC/EUREMA_PB
09:31:55 Agent Paused SV on BTC/EUR: daily-loss cap reached. BTC/EURSV
09:30:14 Agent Watching Donchian channel on SOL/EUR: break forming. SOL/EURDonchian
05 · The prompt

Your strategy is a prompt. Run it on a real engine.

You write the strategy in plain language. The agent runs it on your own Anthropic API key. The engine executes under your caps, on your exchange keys, journal-first.

  • Your model, your API key, held in memory only.
  • Journal-only by default. Dispatch live when you trust it.
  • Caps enforced in the execution path, not in a settings page.
Strategy prompt SV Draft
Trade BTC/EUR on the 5-minute chart.
Enter only on pullbacks in the trend
direction. Stay out when the spread
eats the edge. Respect every cap.
Journal every decision.
Claude · your API key
06 · Pricing

Three tiers. Real prices.

Every tier starts with a 30-day free trial.

Basic

$20/mo
  • Full execution list — every strategy (SV, EMA21, Donchian & more)
  • All exchanges — Binance, Kraken, CoinBase & Bybit Spot
  • Live market data, charts & indicators
  • Per-trade controls: cooldown, profit target, daily cap
Start 30-day trial

30-day free trial included.

Pro

Bespoke
$100/mo
  • Everything in Plus
  • Your custom execution, built by us & added to Claude
  • We scope your strategy from your requirements
Start 30-day trial

30-day free trial included.

Journal-only by default. Dispatch live when you trust it.

Risk disclosure Trading cryptocurrency involves substantial risk of loss. FluxGeko provides execution infrastructure and does not provide investment advice. Past market behavior does not predict future results.