CHEN·R1/mini
instance 0x3F7A · region us-east
Uptime 00:00:00
Cycles 0
PnL · 24h + 7.51%
Drawdown 0.84%
Latency 42ms
Inference active
02 Inference log streaming
› chen@r1
↵ run · ESC disconnect
whitepaper / abstract
Whitepaper · v1.0 · April 2026

An autonomous agent for on-chain alpha.

Chen is a self-improving AI trading agent deployed on BNB Chain. It ingests on-chain telemetry, orderbook microstructure, and social sentiment — then writes, tests, and deploys its own strategies under a hard risk envelope. Access is permissioned: only $CHEN holders staked into the FLAP vault are admitted each epoch.

Authored byChen Labs
NetworkBNB Chain
Citechen.wp/v1.0
01
Background

The retail alpha drain

Retail participants in on-chain markets are structurally disadvantaged. Sophisticated actors run co-located bots, proprietary data feeds, and ML inference pipelines that operate at speeds and scales no individual can replicate. The result is a persistent alpha drain from retail to infrastructure.

The edge in modern markets isn't in the trade — it's in the inference cycle that produced it.
— Chen Labs, design memo · 2025

Chen was built to close this gap. It treats inference as the asset and access to inference as the right earned through participation. Holders who actively stake and claim get the agent's output — not a feed, but the model itself, running, reasoning, and executing in real time.

02
System

Architecture, in five layers

Chen is a closed loop: perceive → reason → bound → execute → learn. Each layer is sandboxed; the risk envelope cannot be overridden by inference. The full cycle settles every 400 milliseconds.

fig 02 · System diagram chen.r1-mini
I
Perception
Mempool feeds, DEX orderbooks, cross-chain sentiment APIs — normalised into a unified feature vector every 400ms.
ingest
II
Inference · r1-mini
A fine-tuned reasoning model evaluates the feature vector against a strategy library, parameterises a candidate, and stress-tests it in simulation before commit.
model
III
Risk envelope
Hard contract: max exposure 4% of vault per trade, trailing stop at 1.2%, max daily drawdown 8%. Cannot be overridden.
contract
IV
Execution & settlement
Approved orders route through aggregated BSC liquidity via private RPC cluster. PnL settles on-chain to FLAP stakers each epoch.
settle
V
Self-improvement loop
After each epoch, Chen prunes underperforming strategies and fine-tunes weights against the epoch's execution trace. No human in the loop.
learn
400ms
Cycle time
Full perceive→execute loop, including simulation pass.
4%
Max exposure
Hard cap per trade. Set by contract, not by model.
7d
Epoch length
Strategy pruning & PnL settlement cadence.
03
Economics

$CHEN — access, share, govern

$CHEN is the access and governance token of the Chen ecosystem. Holding a minimum of 300,000 $CHEN and staking into the FLAP vault entitles holders to inference access, a share of realised PnL, and governance rights over the risk envelope.

Supply allocation · 1,000,000,000 $CHEN
FLAP vault rewards 42%
Liquidity 22%
Treasury 18%
Team (vested 24m) 10%
Community 8%

Holder rights

Agent access · Claim vault rewards each epoch to unlock Chen's live inference dashboard, signal feed, and execution logs.

PnL share · A portion of Chen's realised trading profits is distributed to FLAP vault stakers proportional to stake weight.

Governance · Stakers vote on risk envelope parameters, strategy scope, and protocol upgrades each epoch.

04
Trajectory

Roadmap

✓ Live Q1 · 2026
Phase 1 — Foundation
Core inference engine (chen/r1-mini) deployed
BNB Chain integration & private RPC cluster
FLAP vault contract live, $CHEN token launched
Next Q3 · 2026
Phase 2 — Signal
Multi-chain expansion (Solana, Base)
Social sentiment model v2
Holder dashboard with live PnL attribution
Planned Q1 · 2027
Phase 3 — Scale
On-chain strategy marketplace
Permissioned API for institutional stakers
Cross-chain vault bridging
Risk notice This document is for informational purposes only and does not constitute financial advice. Participation in the FLAP vault and trading activities carry inherent risk including loss of principal. Past performance of the Chen agent does not guarantee future results. Smart-contract risk applies. Consult a licensed advisor before participating.
docs / getting started / overview
01 · Getting started

Access protocol

Chen is a permissioned autonomous trading agent. Access to its inference suite is not open — it is earned through participation in the FLAP vault. Only active $CHEN holders who claim their vault rewards are granted agent access each epoch.

⌖ Why permissioned
Chen's edge depends on a limited circle of execution. Open access would saturate the strategies it discovers; gated access preserves their alpha for active stakers.

02 FLAP vault

The FLAP vault is the access primitive. It accepts $CHEN deposits, accrues rewards each epoch, and gates the agent's inference dashboard behind a snapshot + claim mechanism.

⬡
FLAP vault
access primitive · token-gated
Premium
Minimum stake
300,000$CHEN
Snapshot taken at each epoch boundary.
Unlock trigger
Claim vault rewards
A claim tx signals active participation.
Access window
1epoch
≈ 7 days, refreshed by claim.
Reward source
Agent PnL
Pro-rata to stake weight.
› Non-claimers are queued — the agent serves active stakers first. If claims drop, queue depth grows and access slows.

03 Access lifecycle

Four steps. Each surfaces a verifiable on-chain event.

i
Acquire $CHEN tx
Hold a minimum of 300,000 $CHEN tokens in your connected wallet. Snapshot is taken at each epoch boundary.
ii
Stake into FLAP vault tx
Deposit into the FLAP vault contract. Rewards begin accruing immediately at the epoch rate.
›flap.deposit(amount, lockEpochs)
iii
Claim rewards → unlock access tx
Each claim signals active participation. The agent verifies your claim on-chain and opens a full inference session for the current epoch.
›flap.claim()
iv
Launch agent UI
Once access is verified, hit Launch Agent to enter the live inference dashboard. Signals, logs, and execution stream in real time.

04 Agent commands

Once inside the dashboard, the command bar accepts terse commands. Press ↵ to run, ESC to disconnect.

›
helpread
List all available commands.
›
statusread
Print agent uptime, cycle count, and operational state.
›
riskread
Inspect the current risk envelope and live drawdown.
›
pause / resumeact
Pause inference (positions held) or resume the active loop. Available to governance roles only.
›
clearui
Clear the local log stream view (does not affect on-chain history).

05 Frequently asked

Q.What if I miss claiming an epoch?
Your stake remains and continues to accrue, but agent access is paused until your next claim. Rewards do not expire — they compound until claimed.
Q.Can the agent override its risk envelope?
No. The envelope is a smart-contract constraint, not a model parameter. Inference can request a position only within bounds; out-of-envelope requests are rejected pre-execution.
Q.Where does PnL come from?
Realised on-chain trading profit, net of gas and fees. PnL is settled at epoch close and distributed to active FLAP stakers proportional to stake weight.
Q.Is there a queue?
Yes. Active claimers are served first. Non-claimers are queued in order of stake age. Most epochs serve every active staker; high-load epochs may queue some.
SYS://CHEN.AGENT — V0.4.1 — UPLINK STABLE
N40°·W73° / LAT.OPS
Chen mark
Chen
AGENT ONLINE · INFERENCE STREAM ACTIVE

Autonomous, self-improving AI agent that turns market noise into signal.

Chen runs 24/7 in a sandboxed cloud node. It ingests on-chain telemetry, social sentiment, and orderbook microstructure — then writes, tests, and deploys its own strategies under a hard risk envelope. Access requires holding ≥ 300,000 $CHEN and claiming your FLAP vault rewards each epoch.

README.md agent.ts
1import { runAgent } from '@chen/core'; 2 3const signal = await runAgent({ 4 model: 'chen/r1-mini', 5 scope: ['BSC', 'BNB.chain'], 6 risk: { max: 0.04, stop: 0.012 }, 7}); 8 9// → "rotation forming · attention spike +218%" 10console.log(signal.thesis);