You define the mandate
Choose the capital budget, permitted market, risk limits and when authorization expires.
Planned product capabilityAI CRYPTO TRADING AGENT · PRE-RELEASE
We’re building an AI agent for BTC spot trading. It watches the market around the clock, makes its own buy, sell, hold and sizing decisions inside limits you set, and explains every one.
In development. Today’s demo is a synthetic, rule-based simulation.
No LLM decisions or real-money trading are available.
02 / DECIDE
Intended architecture. AI and live execution are not implemented.
THE PRODUCT WE’RE BUILDING
Set the boundaries, then let the intended agent choose what to do within them. The goal is an autonomous decision-maker, not a tool that waits for you to supply a strategy.
Choose the capital budget, permitted market, risk limits and when authorization expires.
Planned product capabilityAnalyze permitted, timestamped information. Propose buy, sell or hold, a position size and an explanation.
AI engine under developmentA separate deterministic layer checks permissions and risk before any intended execution.
Basic checks exist in simulationDecision quality has to be earned. We plan to compare LLM, quantitative and hybrid approaches against simple benchmarks after costs. We have no verified alpha or evidence of profitability. If the agent ever progresses to real-money use, we will first publish its forward paper-trading record, losses included. Paper results are simulated: they are not real returns and do not predict future performance.
AN HONEST LOOK AT TODAY
This replay uses the existing simulator, unchanged. Seeded fictional prices and synthetic news fixtures drive fixed rules. No LLM generates these decisions.
FICTIONAL BTC PRICE · INDEXED TO 100
1,080 synthetic hourly bars · 48-bar warm-up · 10 bps fee + 5 bps slippage per fill. These are simulation assumptions, not exchange quotes.
DECISION LOG · SIMULATED ORDER INTENTS
| Fictional time (UTC) | Intent | Target | Outcome | Why |
|---|
The Trend, News and Risk modules are deterministic rules. “News” is a synthetic fixture, not a news service. Signals use prior completed bars; simulated fills use the next open. The replay shows the guarded strategy only.
Basic checks cover new exposure, daily loss, buy frequency, stale data, a kill switch and declared sources. They cannot guarantee maximum losses. This demo has no order book, network latency, live paper service, model inference or exchange integration.
Identical engine version, seed and settings reproduce identical outputs. The JSON includes the synthetic inputs, their SHA-256 hash, settings and full result. Synthetic results cannot establish investment usefulness or profitability.
ONE AGENT. TWO WAYS TO USE IT.
We’re testing demand before choosing the MVP. Both proposed offers originate their own trading decisions and share mandatory permissions and risk controls.
$49 / month planned price
A focused autonomous agent, with a mandate you control and understandable explanations of its decisions.
$149 / month per team planned price
The same autonomous agent, with a minimal layer for teams that need to govern deployment and integrate decisions.
Prices are the planned launch prices we’re testing, for the same agent in both offers. Nothing is available or charged today, and no exchange connection is requested.
BUILD WITH US
Tell us what you would delegate, what you would need to trust, and whether you would pay for it.
Join the research and early-access list. We may invite you to an interview or a clearly scoped paper pilot. Signing up does not provide trading access.
No deposit. No exchange credentials. No promised launch date.
A FEW CLEAR ANSWERS
No. Leeboard is pre-release. The current prototype is a rule-based simulator. A genuine AI engine, continuous paper operation and production exchange execution are proposed development phases.
That is the intended product direction: customers authorize autonomous execution inside an explicit mandate. Live execution would require testing, a security review, a compliance assessment and separate authorization. This website cannot place trades or connect accounts.
No. It uses synthetic prices and fixed rules, without an LLM. Neither the demo nor the research establishes profitability or verified alpha. Real trading can lose capital; risk limits do not guarantee a maximum loss.
We’re researching whether teams would pay more for governance and deployment capabilities around the same agent. Team roles, approvals and API access are proposed, not implemented. A signup helps us decide what is worth building.