Why we’re building the trading app with an agent
16 August 2026 · Updated 26 September 2026 · 6 min read
In short
- Sparkling is the trading app with an agent: swaps, perps, prediction markets and tokenized stocks from one self-custodial balance.
- The agent does the work – research, sizing, venue choice, simulation. You make the decision, with one button.
- Agentic execution became a commodity in 2026. Being trusted with the decision did not.
- Your keys never leave your device – every action stays a proposal until you sign it.
Almost any wallet or broker can now turn a plain-language instruction into a transaction. That was the hard problem a year ago; it is table stakes today. What it leaves unanswered is the question that actually decides whether you press the button.
Execution stopped being the hard part
Through 2026 nearly every major wallet and broker shipped some form of AI-assisted execution. Ask in plain language, get a transaction. That capability is now table stakes: it ships, it works, and it is available in a dozen apps.
What none of it settled is the harder question – why you should trust the thing on the other side of the prompt. The common answer was to hand the risk back to the user: confirm each order yourself, monitor the agent yourself, and accept that mistakes are yours. The button became a liability shield rather than a product.
Autonomy is the wrong answer
The case for fully autonomous trading agents got weaker every month of 2026. Public experiments in letting a model trade unsupervised ended in heavy drawdowns. Single mis-parsed values produced six-figure losses. Agent tool ecosystems turned out to be a live supply-chain surface, with malicious skills reaching a meaningful share of some public catalogues.
The lesson is not that agents are useless. It is that the split matters: an agent is excellent at preparation and unreliable at final judgement with money at risk. So we split it exactly there. The agent does the research, the sizing, the venue comparison and the simulation. You make one hundred percent of the decisions.
What our confirm actually is
A signature request is not a decision. It shows you a payload and asks you to trust it. Ours shows the thesis, the size and why that size, the venues compared and the one chosen, the simulated outcome, and what you stand to lose. One card, one button.
That card is the product. Everything else – the routing, the market coverage, the plain-language interface – exists to make the card worth pressing.
Every onchain market from one balance
The coverage is deliberately wide: Hyperliquid perps, with more perp venues such as Lighter on the way, Polymarket prediction markets, spot swaps and tokenized stocks – all from one self-custodial balance. Venue fragmentation is what makes a single access point valuable; as liquidity splits across more places, honest routing is worth more, not less.
Frequently asked questions
What is a trading app with an agent?
It is an app where an AI agent does the preparation for a financial action – research, sizing, venue selection and simulation – and presents it as one decision you approve. The agent does the work; the human makes the call.
Does the agent trade on its own?
No. Every action is a proposal until you confirm it. Sparkling is self-custodial: your keys never leave your device, so nothing can move without your signature.
Which markets can I trade?
Hyperliquid perps, Polymarket prediction markets, spot swaps across chains and tokenized stocks via xStocks – all from one self-custodial balance. More perp venues, starting with Lighter, are coming soon.