The Five Things You Can Ask Cambio's Composer

The chat input at the top of cambio.one is not an open-ended language model. It is not a customer-support widget. It is not a wallet agent. It is a structured intent parser with a friendly conversational surface, and it understands exactly five categories of request. This post walks through all five — with real example phrasings — so you know what works, what doesn't, and what's happening under the hood when you press enter.

If you have not read the anchor post on what 'AI swap' means at Cambio, that one comes first. This post assumes you already know that the AI is a parser plus an explainer, and that deterministic code — not the model — gets the quote.

The five categories at a glance

Each category maps to a different parse family in the backend and routes to a different quote tool. The composer auto-detects which category you mean from the wording of your message, but you can also click a chip to lock the category before you start typing — useful when you want to remove ambiguity.

• Swap — direct execution, amount you want to send (amount-in)
• Execute — reverse direction, amount you want to receive (amount-out)
• Suggest — multi-option exploration ("I have X, what should I do?")
• Compare — cross-network or cross-exchange comparison
• Help — canned educational answers (no quote produced)

1. Swap — the most common path

Swap is what most people are looking for. You have a specific amount of one asset and you want a specific other asset. The composer parses five fields from your message: from-currency, from-network, to-currency, to-network, and amount-in.

Example phrasings the parser recognizes:

• "Swap 0.3 BNB to USDT on BSC"
• "Convert 100 USDT on Tron to USDC on Solana"
• "0.0025 BTC for ETH"
• "Exchange 50 USDT BEP20 to ETH"

When a network is missing or ambiguous, the parser asks. "USDT" alone could be on BSC, Ethereum, Tron, or Solana — so the composer follows up with "which network for USDT?" instead of silently guessing. This is a deliberate design choice: silent guessing is the most common way users lose funds on instant exchanges.

Under the hood, Swap maps to a single quote request: given (from, fromNetwork, to, toNetwork, amountIn), the selected execution route quotes your exact amount, the network fee is stated, and dated reference figures for other exchanges are attached. The AI never picks the route or the price.

2. Execute — when you know the receive amount, not the send amount

Execute is the reverse of Swap. You know how much you want to receive on the other side — typically because you are paying a specific invoice, settling a specific bill, or topping up to a target balance. The composer accepts amount-out language and solves the inverse: how much from-side do you need to send to receive that amount.

Example phrasings:

• "Swap enough ETH to get 500 USDT after fees"
• "How much BNB do I need to send to receive 200 USDT on BSC?"
• "I need exactly 1 ETH — what does that cost in BTC?"

Execute is the same path under the hood, just inverted: Cambio works out the send amount that produces the receive amount you asked for, after the network fee. The same five fields are parsed; the only difference is which side is the constraint.

A subtle but important detail: Execute applies the route's limits on the send side, not the receive side. If the send amount needed is outside what the route accepts right now, the composer says so before any deposit address exists, and shows you the largest amount it can do.

3. Suggest — multi-option exploration

Suggest is for the moment when you have not decided yet. You know what you have, but you have not committed to what you want. The composer takes an open-ended portfolio-shaped question and returns a short list of options, explained plainly (not an aggressive recommendation).

Example phrasings:

• "I have $500 in BNB — what's a low-fee way into stables?"
• "How do I move ETH off Ethereum to a chain with lower fees?"
• "I have leftover SOL — what can I do with it?"

Suggest does something the other categories do not: it prices a few candidate swaps and returns a short list — usually three options. Each option is a fully-priced quote, not a hypothetical. The user picks one, and that pick becomes a normal Swap.

This is the category most likely to feel like a conversation. The AI explains the differences in plain language — which network has lower fees, which stablecoin you would end up holding — and leaves the numbers to the quote cards. The selection logic is still deterministic. The AI just narrates it.

4. Compare — cross-network or cross-exchange

Compare shows how a Cambio quote sits beside other exchanges. Sometimes you do not want to execute — you want to read the table. The composer recognizes comparison-shaped questions and returns the comparison on its own, with no committed route.

Example phrasings:

• "Compare USDT prices across all networks"
• "How does Cambio's quote for 1 ETH compare with other exchanges?"
• "What would other exchanges give for 0.5 BNB to USDT?"

The rows beside Cambio's quote in Compare are the same dated reference figures shown with every quote: checks of other exchanges' own websites with the time of each check, or Est. rows with their dates. Compare just removes the "ready to swap" frame and shows you the table on its own. When you decide, you can pivot to a Swap from the same composer thread without retyping the pair.

5. Help — the only category that does not place a trade

Help is the exception. It is the only category that does not produce a quote. Instead, it returns canned educational answers to a small set of recognized questions about how Cambio works.

Example phrasings:

• "How is the network fee shown?"
• "What are the limits for a pair?"
• "How long does a swap take?"
• "What chains do you support?"

The answers here are not generated freshly each time. They are written, reviewed, and stored as fixed strings — the same text appears for every user who asks the same question. This is intentional. We do not want the model improvising answers to questions about caps, fees, or supported chains, because improvised answers are exactly how policy claims drift. The FAQ page is also where these answers live in long form.

What about everything else?

If your message does not match any of the five categories, the composer says so. It does not invent a category. It does not silently fall back to "let me try anyway and see what happens." It responds with a short message — something like "I can help with swap, execute, suggest, compare, or help — could you rephrase?" — and offers to expand a chip.

The most common cause of an unrecognized message is a missing network. "Swap 100 USDT to BTC" parses cleanly except that USDT is ambiguous across four networks. The composer asks. The user clarifies. The trade proceeds.

The second most common cause is an unsupported pair. Cambio quotes 118 tokens across 25 networks, and a pair outside what its execution routes serve is refused. We tell you immediately — no surprise mid-flow.

Why exactly five, and not "anything you can imagine"

Bounded scope is a feature. Most natural-language interfaces that promise 'ask me anything' fail two ways: they over-promise on edge cases (asking a model to interpret a question it has no policy for produces unpredictable answers) and they make the user uncertain about what is and is not a valid request. The user starts hedging their prompts. The trust frontier blurs.

Five categories is a small enough surface that we can write deterministic policy for each one, test every parse path, and tell you exactly what to expect. It is also large enough to cover the actual jobs people use the composer for. We have not yet found a real swap use case that does not fall into one of these five.

If we ever do find one — and the request pattern is common enough to deserve its own first-class parse family — we will add a sixth category, write the policy, and document it. We will not silently widen the parser's scope.

Try it

The fastest way to internalize all of this is to type any of the example phrasings above into the composer on the homepage. The composer will tell you which category it identified, and the resulting quote card will show you exactly what the deterministic router did with your intent.

The next post in the series explains the reference figures beside every quote — where they come from, why they are dated rather than live, and why we publish the raw rows.