Three types, two modes
Every agent is a pipeline: a chain of steps that watches the chain, filters noise, and either alerts you or trades for you. You build them visually in the pipeline builder. No code.
Pick your pipeline
Follow Wallet
Watch a wallet. When it buys or sells, your agent copies the trade or sends you a signal
New Tokens
Scan every new Uniswap and Uniswap pool. Filter by safety checks, snipe the ones that pass
Token Trading
Watch a specific token. Set triggers for price drops, volume spikes, or market cap thresholds. Trade or alert
Signal vs Trading mode
Signal mode
Alerts only. No ETH moves. You get notified when conditions match and decide what to do yourself
Trading mode
Real execution. Swaps happen automatically via 0x with Flashbots bundles for MEV protection. ETH from your wallet
You can switch modes at any time from the agent page. Trading mode requires a funded platform wallet.
The four steps
Every pipeline is four connected nodes. You configure each one in the builder:
- 01SOURCE
Source
Where the data comes from. A wallet address, a pool scanner, or a token address
- 02FILTER
Filter
Rules that decide what gets through. Trade size, LP depth, contract flags, win rate, and more
- 03EXECUTE
Execute
In signal mode: nothing happens here, the alert goes straight to notify. In trading mode: 0x swap with Flashbots MEV protection
- 04NOTIFY
Notify
Signal lands in your inbox. If Telegram is connected, it pushes to your phone too
Every trade goes through an 11-rule safety gate before execution: daily caps, duplicate checks, slippage limits, liquidity depth, and more. If any rule fails, the trade is blocked and you get a notification explaining why
Agent sub-wallets
Every trading agent runs on its own dedicated Ethereum wallet, separate from your main platform wallet. This is the blast-radius cap: an agent that misfires can only spend what's on its own sub-wallet.
Auto-created on deploy
The moment you hit Deploy on a trading agent, a fresh Ethereum keypair is generated and encrypted at rest. You don't have to set anything up — the address shows up on the agent page.
Funded from your main
Use the Allocate button on the agent page to move ETH from your main platform wallet to the sub-wallet. Set the amount — that's your hard cap on what this agent can ever spend.
Visible balance + PnL
The agent page shows live ETH balance, open positions (with mark-to-market), realized PnL, and the full trade log. Same data also rolls up into Portfolio so you see everything in one place.
Withdraw anytime
Hit Withdraw on the agent page to move ETH back to your main wallet. Open positions can either be closed first (clean exit) or transferred as-is (you keep them on main). Pausing the agent doesn't auto-withdraw — you do that explicitly.
Export key for self-custody
Each sub-wallet's private key can be exported to MetaMask independently of your main wallet. If you export a sub-wallet, the agent stops being able to sign for it — you become the sole owner. Useful when you graduate an agent and want to take its winnings off-platform.
Delete consolidates back
Deleting an agent confirms a one-time sweep: any remaining ETH plus open positions move back to your main wallet, then the sub-wallet is decommissioned. No funds are stranded.
Common questions
Can an agent steal my funds?+
Each trading agent runs on its own sub-wallet with a hard allocation you set yourself — that's the blast-radius cap. The agent code only ever issues 0x swap transactions, never arbitrary outgoing transfers. Worst case if an agent strategy goes wrong: the allocation on that sub-wallet gets traded into bad tokens. Your main balance is untouched. We're custodial overall (see Portfolio docs), so the relationship is "trust the platform" not "trust the agent."
Does an LLM decide every trade?+
Only the ones you configure that way. Filter nodes are deterministic code: if a wallet buys more than 0.5 ETH of a pool with $10k+ liquidity and no mint authority, the trade fires. No AI in that path. Separately, a Token Analyst node can ask an LLM for a qualitative take on a mint before the trade fires. That is the only place credits burn per-run and the only place latency creeps up.
How do I know what an agent just did?+
Every agent has a Decisions log on its page. Every signal it emitted, every trade it fired, every rule that blocked a trade, every credit it burned, timestamped and linked to the source event on-chain. Nothing happens invisibly. If something looks wrong, the log tells you which filter approved it.
What stops an agent from draining my allocation in a night?+
Daily cap on ETH spent, max trades per day, per-trade size cap, and a duplicate-guard that blocks the same token from firing twice inside a cooldown window. Any of those tripping pauses the agent and sends you a notification. You set the numbers, the gate enforces them.
Do I pay the creator when I follow their agent?+
Follow is free. You only pay the per-signal fee when a signal actually fires (0.5 credits by default, around half a cent). Half of that fee goes to the creator as revenue, 25% comes back to you as cashback credits. Your trading profit on the actual swap belongs to you in full.
What happens when an agent dies or I pause it?+
Scheduler stops calling that agent's pipeline immediately. Open positions stay open (take-profit and stop-loss keep running). The sub-wallet still holds whatever it bought. Resume the agent and it picks up where it left off. Delete it and the sub-wallet consolidates back to your main balance after you confirm.