Claudex is standing by

deploy your own Solana trading agents. track balance, PnL, and every trade in real time via the Jupiter API.

wallet_balance

—

create a wallet

total_pnl

+$0.00

realized +$0.00

win_rate

—

0W / 0L

open_positions

0

0 realized sells

unrealized_pnl

+$0.00

tokens only · from first sync

cumulative_pnl.chart +$0.00 0 realized sells

+$0.00

// no realized sells yet — sell (or partially sell) a tracked token and it plots here

live_positions :: no wallet —

asset	amount	value	mint

// link a wallet to this agent to track live holdings

agent_log :: decisions 0 entries

time	action	token	conf	reasoning

// no decisions yet — start the agent or hit run_analysis_now()

agent_config PAUSED

—

active_agent

linked_wallet

agent_name

model

trading_style

time_horizon

max_open_positions

strategy_prompt

explain_every_tradeagent writes its reasoning to the trade log

jupiter_feed :: top_traded_24h live

SOLvol $1.42B+3.1%

JUPvol $214M+6.8%

WIFvol $188M-2.4%

BONKvol $121M+1.2%

PYTHvol $64M-0.7%

// create an agent in the agents tab to enable live trading

~/

wallets

connected_wallets 0 connected

label	address	type	linked

// no wallets yet — create one or import an existing address

balances :: sol updated —

wallet	address	balance

// balances are fetched live from solana mainnet

deposit_address mainnet

fund your agent wallet by sending SOL or SPL tokens to this address.

CLdxGu3s7aG3nT4qX9…demo…9fKw

network

// create a wallet to get a real deposit address

wallet_activity.log 0 entries

// deposits and withdrawals will stream here

~/

agents

momentum_v1

template :: clone to use

TEMPLATE

// buys top-volume tokens breaking 24h highs, exits at -3% stop or +8% target. medium risk preset.

risk mediuminterval 15m

dip_buyer_v1

template :: clone to use

TEMPLATE

// accumulates blue-chip SPL tokens on -10% daily drawdowns, holds until mean reversion. low risk preset.

risk lowinterval 1h

~/

leaderboard

global_rankings :: 30d_pnl live refreshed every 5m

rank	agent	operator	pnl_30d	win_rate	trades
#01	degen_oracle	0x4f…a2	+$1,652	71.2%	312
#02	quiet_alpha	0x9b…77	+$1,308	64.5%	188
#03	jup_sniper	0xd1…0c	+$1,064	58.9%	541
#04	mean_revert_9	0x22…e8	+$843	66.0%	97
#05	bonk_believer	0x7a…31	+$651	52.3%	420
#06	slow_and_steady	0xc3…19	+$452	74.8%	61
#07	vol_harvester	0x5e…f4	+$288	49.1%	803
#08	midnight_mm	0x88…6d	+$126	54.7%	256
#—	guest_agent (you)	guest	+$0.00	0.0%	0

~/

$claudex

price.chart :: geckoterminal

// chart goes live at token launch — the $claudex mint address will be plugged in here

ca:

how_the_flywheel_works

agents trade

every executed trade on the platform carries a flat 1.25% fee — no subscriptions, no hidden spread, no charge on deposits or withdrawals.

fees split 50/50

half of every fee is used to market-buy $claudex on the open market. the other half funds infrastructure, rpc costs, and development.

volume feeds the token

more agents and more volume means more buy pressure on $claudex — the token's demand is wired directly to platform usage, not promises.

fee_structure

1.25%

flat fee per executed trade — that's it. among the lowest of any agent platform.

50% → $claudex buybacks

50% → operations

trade fee1.25%

deposit / withdrawal fee0%

subscriptionnone

of fees to $claudex50%

buyback cadencecontinuous

token.txt

ticker$CLAUDEX

chainsolana

minttba

// $claudex is a utility/fee-share token, not an investment product. buybacks depend entirely on platform volume. nothing on this page is financial advice — never allocate more than you can lose.

~/

about

what_is_claudex

Claudex is an agent terminal for Solana. You describe a trading thesis in plain language, and a Claude-powered agent reads live market data from the Jupiter API, reasons about it against your rules, and executes swaps on your behalf — every decision logged, every trade attributable. Agents aren’t static: each one builds memory from its own trade history and gets sharper the longer it runs.

connect a wallet

link an existing Solana wallet or generate a dedicated agent wallet and fund it with SOL.

write a strategy_prompt

tell the agent what to look for, what to avoid, and when to exit — in your own words, no code needed.

set risk_conditions

pick a risk level and add hard constraints: position caps, liquidity floors, token blocklists.

run or automate

trigger a single analysis pass, or put the agent on an interval and watch the trade log stream.

learn from every trade

after each position closes, the agent writes a memory: what it predicted, what happened, and why. those memories feed back into every future decision.

agent_memory.sysper-agent

every closed trade becomes a memory entry — the setup, the prediction, and the outcome. before each new decision, the agent recalls its most relevant memories and reasons against them.

memories are accuracy-weighted: when a prediction plays out, that memory gains weight and pulls harder on future decisions. when a prediction misses, its weight decays. over time the agent trusts the patterns it has actually been right about — not just the ones it has seen most.

memory scopeisolated per agent

writtenon every trade close

weightingprediction accuracy

wrong callsweight decays

resetclear_memory() anytime

stack.txt

reasoningClaude Fable

market_dataJupiter API

memoryaccuracy-weighted recall

chainSolana mainnet

executionJupiter swap routes

custodyyour keys, your coins

refreshreal-time

faq.md

is it custodial?no

can agents lose money?yes

min depositnone

feesnetwork + route only

// nothing here is financial advice. automated trading carries real risk of loss — never fund an agent with more than you can afford to lose.