Perps without tears.
Agent NRKT is your lightning-fast AI signal + execution bot for 
Hyperliquid and 
Aster.
ML + Reasoning + Vision — One Decision
We fuse traditional ML, reasoning LLMs, and vision chart‑readers into one ensemble. Each model reviews the same context and charts; tiered consensus turns agreement into a single, auditable trade plan. The same vision‑first algorithm that makes TrenchBuddy early on memecoins.
How it works
- •Traditional ML features and pattern stats set the stage.
- •Reasoning LLMs weigh narrative and context across timeframes.
- •Vision models read charts and spot structure, levels, and momentum.
- •Tiered rules pick direction and size: Consensus (T1), Specialists (T2), Scalps (T3).
- •Full audit: prompts, images, responses and votes are logged for transparency.
Preview calls from the ensemble
BTC-PERP
SOL-PERP
DOGE-PERP
Preview
Fast when it flies. Careful when it whipsaws.
Orders are idempotent, sized to venue rules, retried with jitter, and guarded by a kill switch.
Order Lifecycle
Autonomous trading execution flow with real-time progress
Signal
Analyzing market signals and patterns using ensemble prediction models
TradePlan Anatomy
- • Idempotent placement
- • Venue tick/lot conform
- • Kill‑switch armed
- • Correlation cap: 0.35
- • Ensemble T1 consensus 78% (3/4 models agree)
- • Funding normalizing; sweep of local liquidity
- • Momentum divergence resolving on 15m → 1h
Positions, delta, and health (preview)
Preview metrics • Derived from Genesis cohort simulation
Pending approvals (preview)
3The approval queue validates every trade before it fires. Here’s the simulated flow we’ll ship with the beta cohort.
“No action” is a decision
We believe restraint is alpha. The engine prefers a clean “no action” over forced trades in chop.
“No Action” Wins
Illustrative spark charts, not market data.
- Consensus not metIf models disagree or confidence is low, we stand down.
- Mixed signalsConflicting timeframes or invalidation → no trade.
- GuardrailsFlip throttles and correlation caps keep discipline.
Supported venues
Hyperliquid
DEXThe OG perps dex of the cycle.
Aster
DEXThe up-and-comer, backed by CZ.
Paper
SandboxPretend to trade, then cry because you didn't actually get that 50x.
Simple, fair, capacity‑based
Stake TRENCH to secure your place on the waitlist. The staking mechanism will be the same system that powers staking on the OG TrenchBuddy website — almost finalized as we work through the final legal nuances. Bigger stakers get priority, early holders get a head start . If you unstake or sell, you'll lose your position.
Stake TRENCH for priority access
Staking your TRENCH tokens secures your waitlist position. The staking system mirrors the OG TrenchBuddy website mechanism — ensuring consistent utility across platforms. Everyone else is ranked by staked amount × retro head start. We're finalizing legal frameworks to ensure compliance.
Priority Queue
Fair access based on contribution & commitment
Staking requirement: ≥100,000 TRENCH tokens. Retro boost: +5.0% for holders ≥3 months. Same staking system as OG TrenchBuddy website — ensuring unified utility. Legal compliance measures being finalized.
Staking-based priority system
The staking mechanism for waitlist priority will be the exact same system that powers staking on the OG TrenchBuddy website. This ensures consistent utility and value across the entire TrenchBuddy ecosystem. We're in the final stages of establishing the legal framework to ensure full compliance.
- • Same staking contracts and mechanisms as OG TrenchBuddy
- • Unified token utility across all TrenchBuddy platforms
- • Legal compliance framework being finalized
- • Cross-platform staking rewards and benefits
We wish everyone could get a spot on day one. To keep reliability and safety high, we invite people as fast as humanly possible while we scale dedicated capacity.
- • Per‑user ML ranking calculates your position from on‑chain signals.
- • Per‑user execution: each user runs on a separately hosted execution service — no shared queue — so your trades never wait behind someone else's.
- • Provisioning and monitoring isolated services takes real infrastructure; invites per wave increase as capacity grows.
Why there’s a wait (and why it’s worth it)
We wish this could be free and unlimited. But giving every user their own high‑performance execution service costs real infrastructure. The upside: speed, reliability, and zero waiting behind other people’s trades.
Why per‑user is better
Shared bots make everyone wait on the same queue. We give you your own lane.
Per‑User Isolation
Contention under load
No waiting behind others
Hey, what gives? Why isn’t this open to all right now?
Short answer: dedicated horsepower. Unlike ‘one‑signal‑for‑everyone’ bots that share compute and queues, we spin up a separate execution stack for you and only you. That guarantees trades tailored to your exact settings and the fastest path from signal → order without waiting behind other users.
Why not let everyone in on day one?
We’re onboarding as fast as humanly possible. ML ranking runs per wallet, and each user gets their own execution service. Standing up, isolating, and monitoring those services safely takes real capacity — which we’re scaling aggressively.
Agent NRKT launch sequence in progress
We’re putting the final touches on the production console and invite flow. Drop back in soon for the private beta schedule.