Skip to content
← comparisons

The checklist · 8 min read

Six questions to ask before you connect anything to your money

Every automated trading service answers the same handful of questions, and most of them answer in marketing language that sounds reassuring and settles nothing. "Your funds are safe." "Non-custodial." "Bank-grade security." None of those are answers. They are the absence of an answer, phrased confidently.

Below are the six questions worth putting to any service, what a real answer sounds like, and what the evasive version sounds like. We build Tradevo, so each one also carries our own answer — including the four where something else beats us. A checklist you cannot fail your own product against is not a checklist.

1. Who can move the assets, and whose cooperation is needed to stop it?

A good answer names every party that can move funds, and says plainly whether ending the arrangement requires the operator to do anything. A bad answer says "you stay in control" without saying who else also does.

Ours. Two parties. You: assets sit in a smart account on Base whose owner is your embedded key, a send from inside the app is signed client-side with no call to our API, and the key is exportable at any time. And us: subscribing attaches a delegated session signer to that key.

Ending it needs us. Cancelling is durable and immediate — the schedulers re-read subscription status inside the row lock and abandon the tick. Removing the signer is separate, operator-performed and best-effort, and it deliberately does not fire when another active subscription still relies on it, or when a liquidation leg failed and the tokens still need selling. There is no user-callable revoke.

2. What exactly can that credential do, and who enforces the limit?

The question most services quietly skip. A good answer names the enforcement layer — the chain, the venue, the provider, or the operator's own code — and does not blur them. A bad answer is any sentence of the form "it can only do X" with nothing after it saying who guarantees that.

Ours. The grant is account-level, not action-level. Our own server comment says the provider "authorizes us to sign on behalf of any user wallet that has attached this signer."

Three candidate enforcement layers, honestly:

So the honest sentence is that the bound is ours to keep. Narrower in practice than a key that can trade a whole account — and still a promise rather than a mechanism. Our repository also contains a helper that signs arbitrary calls from the same account, used for cross-chain bridging.

3. If the operator vanished tonight, could you still reach your funds?

A good answer describes the recovery path concretely enough to rehearse. A bad answer is "your funds are always yours" with no mechanics attached.

Ours. Yes, with an order of operations worth knowing before you deposit. The assets are in an ERC-4337 smart account on Base; its address is shown in the app with a block-explorer link, so you can check the balance without us. Its owner is your embedded key, which you can export.

The part people get wrong: exporting hands you the owner key, it does not by itself move anything and it does not remove the delegation. Doing it entirely without us means driving the smart account directly — funding the owner key with ETH and calling the account — not importing into a consumer wallet and pressing send. Doing it inside our app is easier and still rides infrastructure we configure.

4. How would you know if it silently stopped running?

Almost nobody asks this, and it is the most common real failure. A scheduled action that should have happened and did not produces no error and looks exactly like a quiet week.

A good answer points at something you can check yourself. A bad answer is "we monitor it."

Ours, including where it is weaker than it looks. There is an unauthenticated deep-health endpoint anyone can poll without an account, answering five named checks and a timestamp, and returning 503 while any check fails.

Three real limits. The status code alone does not prove we are alive: the endpoint serves the watchdog's last published report with no staleness guard, and a tick exits immediately if the previous one is still marked running — so a single hung tick freezes the report. If the watchdog stops while the API stays up, it can keep answering a cheerful 200. Read the timestamp, not the status code. That is the check we would want a careful reader to make, so it belongs here rather than in a footnote.

5. What does the published record actually prove?

A good answer distinguishes what was simulated from what happened, and labels which is which at the point of decision. A bad answer shows one curve.

Ours. Two separate artifacts, deliberately labelled. Headline statistics on a strategy card are a backtest computed over a trailing window at seed time, every figure labelled with the word backtest, and the line directly above the subscribe button says the figures are hypothetical and not live trading. When price history is missing the seed refuses to print a flat zero and marks the strategy not-backtested instead — which is why 5 of the 20 strategies currently listed carry "EXPERIMENTAL · NOT BACKTESTED" and show no figures at all.

The forward track is a different thing: an append-only daily record, one row per strategy per day. The 15 non-backtested-exempt strategies each hold between 72 and 81 rows, beginning mid-June 2026, with no missing days.

What that proves is narrow and worth stating: that a number was written down each day and not edited afterwards. It does not establish that a pattern continues. Anyone presenting an unbroken record as though it settles the question has overstated their own process.

6. What does it cost to leave?

A good answer covers fees, lockups, notice periods, and what happens to open positions. A bad answer mentions the monthly price and stops.

Ours. No exit fee, no lockup, no notice period, and no per-strategy charge — all 20 currently listed are priced at zero per month. The optional Pro tier is $14.99 a month and you cancel it yourself in Stripe's own billing portal; it is independent of cancelling a strategy in either direction.

If you sell back to USDC on the way out, the costs are third-party ones — the pool fee and slippage within a per-token cap, tightest for majors and wider for thin tokens. Any leg worth under a dollar is skipped as dust and stays in your wallet as tokens rather than being sold.

When another architecture is the better choice

Applying this list to ourselves, four of the six have an honest answer that favours something else:

Where we are genuinely different is narrower than a slogan: the assets are not on a venue's balance sheet, and every execution is a public transaction you can check without asking us.

Related: What a trading bot can do with your exchange API keys · Tradevo's method

Tradevo Technologies builds evidence infrastructure for systematic trading. Nothing on this page is investment advice, and no part of it describes how any strategy performed. Crypto assets are volatile and you can lose your entire allocation.

See the mechanism, not a pitch

How strategies are admitted, what the authorisation permits, and how the forward record is kept.

Tradevo's methodOther comparisons