The Kalshi copy trading bot that never holds your money.

A Kalshi copy trading bot is a system that watches specific traders and takes the same side of the same event contract on your account, automatically, before the price has moved away from theirs. The hard parts are not the idea. They are seeing the trade quickly enough to act on it, deciding whether a rule is worth running before it spends anything, and making sure the thing placing orders can never do more damage than you agreed to.

Autopilot runs on your Kalshi account, not ours

Rivo autopilot is non-custodial. You connect your own Kalshi API key, it is encrypted at rest with AES-256-GCM, and every order autopilot places is placed by your key on your account. There is no Rivo balance to top up, no pooled wallet, and no path by which Rivo can move your money out. Disconnecting the key in settings wipes it and stops execution immediately.

This is the part most copy-trading products get backwards. A bot that requires you to deposit into it has made a custody relationship out of what should be an execution relationship. Kalshi already issues per-user API keys precisely so this is not necessary.

A rule earns a record before it is allowed to spend

Every autopilot rule starts on paper. It fills against real Kalshi trades as they happen, marks open positions against live prices, and is scored properly once those markets resolve. You are looking at what the rule would have done on the trades that actually occurred, not at a curve fitted after the fact.

When the record justifies it, the same rule is handed to live execution without being rewritten. That ordering is deliberate. The number on the button is a measured one, which is the only honest basis for turning real money loose on it. Nothing here is a projection of future returns, and results on paper do not promise results live: entry prices drift, fills are not guaranteed, and a rule can stop working.

What autopilot will and will not do

• Entries only. Autopilot opens positions. It does not mirror a trader out of one, so closing stays your decision. Automatic exits are a separate feature: standing stop-loss, take-profit and trailing rules you arm per position.
• Two ceilings, both yours. A maximum per copy and a maximum per day. Autopilot stops at both. Presets start at $5 per trade against a $25 daily ceiling.
• It skips rather than chases. A trade that has aged past the copy window, or a market whose price has already effectively decided the outcome, is skipped and logged with the reason.
• It stops itself. Three consecutive execution failures pause the rule. Market conditions such as a thin book do not count toward that, so a rule is not retired for the market being difficult.
• Every decision is logged, including the skips. The copy log records why a trade was not copied, which is the only way to answer that question honestly.

Live on Kalshi, coming on Polymarket

Live autopilot executes on Kalshi today. Polymarket is fully covered everywhere else in the product, including the live feed, trader records, alerts, backtests and paper strategies, and live Polymarket execution is still coming. We would rather say that plainly than let a Polymarket rule quietly fail to fire. See the copy-trading guide for how the two venues differ in practice.

Choosing who to copy

A copy bot is only as good as the account behind it, which is why trader selection sits upstream of any of this. Every tracked Kalshi trader has a public record: what we have observed them trade, what resolved, and the profit we computed from those outcomes rather than a number they reported. Start from the Kalshi whale tracker or the biggest wins page, open a profile, and backtest a rule against that trader before you follow them with money.

Driving autopilot from code or an agent

Autopilot is available through the Rivo REST API and the MCP server on the same account and the same quota. Keys carry read, write or trade scopes, live autopilot needs the trade scope plus an explicit confirmation on the call, every order defaults to a dry run, and a per-key daily spend cap fails closed. The developer documentation covers the endpoints, the MCP tools, and the readiness object that names exactly why a copy cannot run.

Backtesting whale activity with strategies

The standard advice in prediction markets is to follow the smart money. It does not say which smart money. Large positions on heavy favourites resolve correctly most of the time and return very little per dollar staked; large positions on longshots lose far more often and pay multiples when they land. Politics, sports and crypto markets behave differently again. Which of those is worth copying is an empirical question, and tracking trades as they happen does not answer it.

A Rivo strategy is that question stated as a rule and answered against data. A rule might read: entry price under 20 cents, trade size over $10,000, new positions only. Every whale trade satisfying it is recorded, and scored win or loss once the underlying market resolves.

A strategy can be tested against history and run forward at the same time. On creation it optionally claims every qualifying trade from the preceding 90 days, most of which sit in markets that have already resolved, so the strategy carries a settled record immediately rather than accumulating one over a season. From that point it also claims new qualifying trades as they print. Positions in markets that are still open are priced against current odds and marked as unrealised; they are reported separately from settled results, and are never blended into a single figure.

Criteria can constrain category, sports league, trade size, the entry price band, the position type (a new position, an addition to an existing one, or a reversal), and free text matched against the market title. Specific markets can also be followed directly, in which case their whale trades join the strategy whether or not they satisfy the other criteria.

Each recorded trade stores the side, entry price and size at the moment it printed, so a strategy's history stays reproducible. Open positions are marked against current odds; settled positions are scored win or loss against the market's resolution. A stake is set per strategy or per trade, so the reported return reflects the size you would have taken rather than the size the whale took, which keeps two strategies comparable instead of dominated by whichever caught the largest single position.

Every strategy reports its own win rate, realized return, open exposure and a cumulative return curve. All figures are paper results that assume a fill at the whale's entry price, and are not investment advice.

Frequently asked questions