01Latency
Information arrives unevenly
Venues and chains do not learn the same price at the same instant. For a short window after something moves, some places are still quoting the old one.
The opportunity
Three things it trades on.
Not a view on where the market is going. Three structural facts about how crypto trades, each of which opens and closes faster than a person can act on it.
02Volatility
Hard moves widen the gaps
The faster prices move, the more often venues disagree and the wider the disagreement. Turbulence that hurts a discretionary book creates openings here.
03Discrepancy
One asset, several prices
The same token is worth different amounts in different places at the same moment. That difference is the trade, and it is taken in full or not at all.
What we will not describe is how any of it is detected or acted on. That is the part worth paying for, and publishing it would retire it.
By hand, versus this
Why nobody does this manually.
Not because traders lack skill. Because the window is shorter than the reaction.
By hand
- You see the gap after it has already started closing.
- One screen, one chain, one pair at a time.
- Sleep, meals and weekends are hours of no coverage.
- Position sizing done in your head, under time pressure.
- What happened last week reconstructed from memory and tabs.
With the agent
- Acted on inside the window, or skipped deliberately.
- Every covered network watched at the same time.
- Continuous. The agent does not sleep or take a weekend.
- Size and loss limits enforced mechanically, every time.
- Every fill on the record with its result and its latency.
Swarms
Many agents, none of them in each other's way.
One agent can only be in one place. So you field many — and each is handed an exclusive share of the market to work. No two agents ever chase the same trade, which means adding agents adds coverage rather than adding collisions.
Exclusive territory01
Each agent owns a share of the search space. Nothing overlaps.
Genuinely parallel02
No queue, no leader, nothing waiting on anything else.
Category coverage03
Swarms spread across strategy types instead of crowding one.
Scales by addition04
More agents means more of the market seen at once.
Pause one, keep the rest05
An agent can be stood down without disturbing its siblings.
Console
Six views. Nothing behind a support ticket.
Built for a phone first, because that is where you will actually check it.
01Overview
What it is doing now
Whether the agent is engaged, what it has made or lost today, and how much room is left inside every limit.
02Swarms
Your fleet
How many agents are live, the share of the market each one holds, and what each has claimed.
03Activity
Every fill
Each trade in sequence with its result, its latency and its venue. Filterable, exportable, nothing withheld.
04Controls
Your limits
Position size, daily loss floor, slippage tolerance, concurrency — and the stop. Changes take effect immediately.
05Wallets
Connected addresses
Which wallets the agent may trade from, their balances, and a one-action disconnect.
06Access
API keys
Scoped credentials for your own systems — individually named, individually revocable, individually logged.
Cadence
A day, as the agent runs it.
00:00
Your limits are already in force
Nothing is decided in the moment that could be decided calmly in advance. Size, loss floor and stop authority are set before anything trades.
—
The swarms work, unattended
Every covered network, continuously. No opportunity waits for you to log in, and nobody has to stay awake to supervise it.
—
Each fill lands on the record
Result, latency, venue and time, written as it happens. Your activity log is never reconstructed afterwards, because it was never incomplete.
—
Limits assert themselves
Hit the daily floor and the agent stands down for the day and tells you why. It has no route to widen its own ceiling to keep going.
23:59
Nothing to reconcile
There is no end-of-day assembly step. The record has been current all day, so your report is simply the record.
Control
Simple controls, because they have to work under pressure.
A control that needs a manual fails in the moment it matters. Engaging, setting limits and stopping are each one unambiguous action — and each is recorded.
Engage / stand down01
One switch. Immediate, from any device.
Maximum position02
The ceiling on capital committed at any moment.
Daily loss floor03
Reach it and the day ends. Not negotiable mid-day.
Slippage tolerance04
How far from quote a fill may land before it is refused.
Emergency stop05
Halts every agent and swarm, unconditionally.
Illustrative. Your console shows your own activity only.
The record
Measured from what settled, not from what was quoted.
Every number shown to you comes from what actually happened on chain — the balance change, the fill, the latency. Not from a model of what should have happened, and never from a backtest standing in for a result.
Live
Freshness
Current the moment you open it.
Exportable
Portability
Take your record with you whenever you want.
Per fill
Granularity
Every action individually, not a daily total.
Append-only
Integrity
Entries are added, never rewritten.
Interoperability
Fits the stack you already run.
Connect the wallet you already use. Sign in through your organisation's identity provider. Pull your activity into your own systems with scoped keys. Nothing here asks you to abandon something that works.
—
Your own wallet
Connect an address you control. The agent trades from it; we never hold the keys to it.
—
Machine credentials
Named, scoped API keys for your own services — each individually revocable and individually logged.
—
Portable records
Your activity exports cleanly. There is no retention advantage in making your data hard to take with you.
Set your limits and let it hunt.
Connect a wallet you control, set the boundaries you are comfortable with, and the agent starts working. You keep the keys and the stop.
Non-custodial. Stoppable at any moment. Trading involves risk of loss.