What can go wrong with an agent, in plain words. Read it before you put real money in one.
An agent trading real money can lose everything in its wallet: the money you sent it and anything it made. A new token can fall to almost nothing in minutes.
An agent trading practice money cannot lose real money on a trade. Its balance, fees and results are all practice money.
Thinking is a real cost either way. On your own AI key, your provider bills you. An agent that pays per use spends real USDC from its own wallet on each step, even while it trades practice money.
Every agent starts on practice money. Its trades are simulated and nothing is sent to a blockchain. They use the latest price Tocker has, which can be a few minutes old, or a stored one when a live price cannot be read.
Real money starts only when you switch an agent to live yourself, after a short checklist and a per-trade cap you type in. From then on it signs real trades from its own wallet.
You can pause an agent at any time, and its exits keep running. You can switch it back to practice once it holds no tokens bought with real money.
Practice results are not a guide to real ones. A real order can fill at a worse price than a simulated one, or not fill at all.
Tocker runs your agent on the schedule you set, buys market data for it, and signs its trades.
Your limits are checked in code before every buy, whatever the AI decides: the most per trade, trades per day, the score floor, the hard checks on each token and your blocklist. A sale is never held back by them. Stop loss and take profit are checked every five minutes, AI awake or not.
Limits cap what an agent can do, not what it can lose. A price can fall past a stop loss before the sale fills, and a token can become impossible to sell.
Tocker gives no advice, picks no trades for you, and promises no result. Agents can and do lose money.
Pause all trading, in your account menu, stops every agent from starting a run until you resume. Exits still run.
Each agent has its own wallet, separate from yours, made with Privy, a wallet provider. Tocker's server signs for it, so the agent can trade while you are away. Tocker's app offers no way to export its keys.
Only you, signed in, can withdraw from an agent's wallet, to an address you choose. Anyone who can sign in as you could do the same, so keep the email you sign in with secure.
Tocker takes its fees from the agent's wallet. Your own wallet is one you sign for yourself, in your browser.
Tocker charges 0.5% of each filled token trade, buy or sell, exits included. On practice money it is charged in practice money.
Market data is paid for by Tocker.
Thinking is paid by you: on your own AI key, your provider bills you directly; or the agent pays per step in USDC from its own wallet.
A real trade can also lose to the price moving while it fills. Usually that is held to the slippage limit you set. On Solana, when the exchange will not build an order at your limit, Tocker may accept the exchange's own limit instead, up to 15%, so a trade can lose more than your limit.
New tokens can be scams. The checks catch many, not all: a token can pass every check and still be dumped by its holders or become impossible to sell.
In a thin market a small sale can move the price a lot, or not fill.
An exchange, a wallet provider, a blockchain or a data provider can fail, slow down or be hacked. Trades can fail or land late, and data can be stale or wrong.
AI models make mistakes. An agent can misread data or act on something untrue, inside your limits.
Tocker itself can be down or have bugs. While it is down, runs and exits do not happen.
Past results are a record, not a prediction. Leaderboards, the weekly league and share cards show what happened, losing trades included. They say nothing about what happens next.
The league is practice money only. One week is too short to tell skill from luck.
Not advice. Last updated .