Market data
- Binance klines
- Kraken OHLC
- Public endpoints
Architecture
How MONUNI decides.
Most tools hand you a signal and leave the hard part to you. MONUNI runs the whole decision — and shows you every stage of it, including the stages where it decided not to trade.
The pipeline
Everything a proposal has to survive.
Data on the left, an order on the right, and five stages in between. Any one of them can end the cycle, and each one records why.
Inputs
Public market data and your own settings. Nothing else.
Account state
- Balances
- Open positions
- Working orders
Your configuration
- Mode & environment
- Risk profile
- Capital share, limits
Features
Computed, not inferred. Under 20 candles it returns nothing rather than a guess.
Derived measurements
- Short & medium momentum
- Relative volume vs average
- ATR % of price
- Range high / low
Agents
Independent strategies, evaluated separately. They do not vote and do not debate.
Core Agent
- Accumulates BTC and ETH on
- pullbacks in an uptrend.
- Never chases.
Momentum Agent
- Buys strength volume confirms.
- Steps aside when BTC is weak.
Each proposal carries
- Confidence 0–100
- Entry, stop, take-profit
- Its reasons, in writing
The gauntlet
Three independent stages. Any one of them can end the cycle.
Allocation
- Sizes from that agent's share
- of the capital you released
Risk engine
- Approve · Reduce · Reject
- Sized from stop distance;
- smallest cap wins
Gates
- Exposure, cooldown, drawdown,
- daily loss, quota, emergency stop
Execution & record
Paper by default. Live only through your own account, and only when you say so.
Paper venue
- Real prices
- Spread, slippage and fees
- charged as a real fill would
Live, via the relay
- Ed25519-signed requests
- One fixed egress address
- Your exchange account
Decision record
- Every decision written down —
- including the refusals
No model in the decision path
The engine is deterministic. The same candles and the same settings produce the same decision, so the reason on the record is the actual reason — not a plausible summary written afterwards.
Exits are handled before entries
Every cycle protects open positions before it looks for new ones. A run that spends its budget opening trades and then cannot manage the ones already on has its priorities backwards.
One run at a time, per connection
A cycle takes a lease before it does anything, and refuses to act on stale market data. Two overlapping runs would be one decision placed twice.
Environment and mode
Two things you control separately.
Environment is whose money is at stake — test or live. Trading mode is how far MONUNI may go. They are deliberately independent, so you can run a real account in Observe for as long as you like.
- 01
Observe
MONUNI analyses and records. It places nothing, and it cannot — the code path that contacts an order API is closed in this mode.
- 02
Paper
Simulated fills against real prices, with real fees and slippage modelled. Still no real order.
- 03
Approval
MONUNI proposes; you decide. Nothing is placed until you say yes, and it re-checks the market before acting on your approval.
- 04
Auto
MONUNI executes within your risk limits without asking each time. It still cannot exceed a single one of them.
What to expect
Most of what it does is decline.
A quiet week with a clear record of why is what a disciplined process looks like, and it is the honest thing to expect. Nobody can promise you a return, and we will not pretend otherwise.