Multi-step DeFi
workflows.
An agent plans them and runs them. Lending, liquidity, hedging, swaps and perpetuals, across four networks, on a clock or at a price. Nothing spends until you start it, and we never hold your money.
Any strategy.
One action.
Pick a request and build it. You read every step before anything happens — solid links carry a step's output into the next; dotted links are the boundaries the engine refuses to cross.
DeFi, as one surface.
What is plugged in today is not what the product is. Nobody waits for us to build their strategy. A protocol integrates once and is available in every loop, including ones written before it existed.
A rule in plain English. It waits, then acts.
In chat, the agent writes a card: where it runs, when it acts, what it does and how much it may spend. You read what it can and can never do, and you approve it with your signature.
- 01
Write it
Say the rule in chat. The agent writes a few lines of English, and a checker refuses any line it cannot turn into a safe plan.
- 02
Test it
Backtest it against simply holding, over real history. Or run it on paper: live prices, simulated fills, no money moves.
- 03
Run it
It acts by itself inside the limits you signed. Stop one card, or all of them, whenever you want.
Your signature covers the card, its limits and its budget, not each trade. A live card spends real money without asking each time, and it can still lose money inside those limits.
How cards work →Perpetuals, from your own account.
Your money stays in your own account at the venue. We hold a trade-only key inside the hardware enclave, and it cannot withdraw to an address of its own.
The agent can still lose money by trading badly inside the limits you set. On Hyperliquid, take-profit and stop-loss can be set when you create an order, not yet on an open position.
About trading venues →Three roles, and only your wallet owns anything.
The agent is a contract. The fee payer pays. Neither can hold a token, and neither was ever given your key. Watch one operation go round.
revokesignature
checkedsubmits ·
pays gassigned op
Your wallet appoints the agent. Only the account can appoint — no admin can point yours at an agent you never chose.
The agent contract calls an approved operation. It has no raw call and no field for a destination, so value can go only to your account or your wallet. It holds no value at any point.
The fee payer submits it and pays the network fee. The hash commits to the exact calldata, so substituted calldata is rejected.
You revoke in one transaction. It is one-way, needs nobody’s permission, and stops the agent at once.
The agent signs inside a hardware enclave.
Nobody at DeFiLoops can read the key out. The enclave signs a step only if it passes a fixed policy, and it does not cap how large a step is. The people who run our cloud account could change the rule that releases the key.
How the enclave works →The questions every operator asks first.
If your question is not here, it is probably the one we want to hear. Bring it to the call.
01Do you ever hold my funds?
No. Your money sits in an account you own, derived from your own address and live at the same address on every network. We never have a key to it, and there is no vault to deposit into.
stuck, not stolen
02How do I recover my funds?
You withdraw them yourself, directly from your account, with your own wallet. You need no agent, no fee payer and no approval from us. The account answers to your signature alone, so recovery is a normal transfer, not a support process.
withdraw any asset, any network, any time
03Do I ever give up control?
Never. You hold unilateral control of the account 100% of the time: you can move, withdraw or unwind everything in it whenever you want, and the permission you granted the agent does not limit any action you take. Revoking the agent needs nobody’s cooperation but your own.
sole owner · unilateral · at all times
04Can it tell me things without spending anything?
Yes, and most of what it can do is this: 5 of its 52 tools spend money and the rest only read. Ask what a price range would have collected day by day on every network and fee tier, how seventeen ways of managing that position would have done over real history, what nine option structures cost today and cost at every roll since 2021, what perpetuals pay on four venues, who is being liquidated and where, and what prediction markets and the rate curve expect. None of it signs anything.
ask before you act
05What exactly can the agent do?
The agent is a contract your account appoints, not a key we hold. It may call one approved operation on one account, and it refuses the rest on chain: it will not target your account directly, will not install permissions or change who owns it, and will not touch an operation the guardian never approved or that you never switched on. Its signing key sits in a hardware enclave. It can still make a bad trade inside those limits: nothing caps a step unless you set a daily limit per token, and the app has no screen for that limit yet.
06How do I stop it?
One transaction from your own wallet stands the agent down. You do not open a ticket, and you do not wait for us. The agent checks the appointment on every call, so it stops at once. Your account can also never be re-pointed at a different agent: the first appointment binds, and the only permitted change is to none at all.
07Who pays gas across all these networks?
A fee payer of ours sends the calls. The platform fee is a markup on the network fee, charged in credits, and it never comes out of the amount a step moves. You never hold the native coin of a network you have not used.
zero signatures to reach a new network
08What happens if a step fails mid-loop?
The plan stops at the failed step and reports the state it stopped in. Steps that have not run do not run. Nothing is left half-approved, and the next scheduled run starts from a clean read of your balances.
09Can I use a protocol you have not integrated?
Not by writing it into a plan, and that is the point. The list of operations is closed, which is what makes it safe to hand to a model: a plan naming anything else is refused when it is written, rather than attempted and failed halfway. Adding to the list is cheap though. A new venue is a new entry, not a change to how plans work, so we can add it in days and your plan does not change shape when we do.
10How do I know what a plan will do before it runs?
Every plan compiles to the board you saw above: each step, the protocol it calls, the chain it lands on, and where the money moves next. You approve that, not a sentence.
11Is this custody, and does it change my reporting?
It is self-custody throughout. Positions sit in an account in your name, so your existing custody, accounting and reporting posture holds — we add execution, not a counterparty.
12What is a strategy card?
A rule in plain English that the agent writes for you in chat. It says where it runs, when it acts, what it does and how much it may spend. A checker refuses any sentence it cannot turn into a safe plan. You read what the card can and can never do, read its backtest, approve it with your signature, and start it.
your signature covers the card and its limits, not each trade
13Can I run a card without risking money?
Yes, as a paper run. An approved card acts on live prices with simulated fills, and no money moves. The kill switch stops a paper run like any other card.
14Do you hold my money on Hyperliquid or Lighter?
No. Hyperliquid and Lighter hold it in your own account at that venue. We hold a trade-only key inside the hardware enclave, and it cannot withdraw to an address of its own. It can still trade badly inside the limits you set.
Already run an agent? Point it here.
Point any MCP client at the address below. It reads the public catalogue without signing in, and your accounts after a wallet sign-in. To run operations from a program, call the paid gateway with a wallet and pay per call.
https://mcp.defiloops.com/mcpBring one strategy you are tired of running by hand.
DeFiLoops is live. Open the app, say the strategy in chat, and read the plan before anything runs.