Every bot on Hyperliquid, from a Telegram bot to a Python script, trades through the same API with the same kind of key. How that key works, which bot fits which market, what goes wrong, and how to choose one.
Updated November 15, 2026
22
min to read
The short answer
A Hyperliquid trading bot is software that trades your Hyperliquid account through its public API. Most use an API wallet: a separate key your main wallet approves, which can place orders but cannot withdraw. Bots run rules such as grids, DCA, signals or copy trading. They follow a strategy without pause, but they can’t make a bad strategy good.
Hyperliquid is an exchange that runs on its own blockchain, best known for perps, short for perpetual futures: contracts with no expiry date that track a price, usually traded with leverage. Every bot on it, whether a hosted grid bot, a Telegram bot or a Python script on your laptop, talks to the same public API. An API (application programming interface) is the door a program uses instead of the website. Hyperliquid’s has two parts. The info endpoint answers questions: prices, order books, candles, and any account’s positions and fills. The exchange endpoint takes signed actions: orders, cancels, leverage changes. A websocket streams the same data live. Nobody gets a faster lane: Hyperliquid says it has no designated market maker program, and no special rebates or latency advantages.
What matters most for your safety is which key signs those actions. Your main wallet can approve an API wallet, which Hyperliquid also calls an agent wallet: a separate key that signs trades for your account without being your main key. Hyperliquid’s official Python SDK describes it, in its agent example, as an agent “that can place trades on behalf of the account” and says it “does not have permission to transfer or withdraw funds.”
| Action | Can an API wallet do it? |
|---|---|
| Place, change and cancel orders | Yes |
| Change leverage and margin | Yes |
| Move money between your own spot and perp balances | Yes, to the same address only |
| Withdraw, or send funds to another address | No |
| Approve another API wallet or a builder fee | No, that takes your main wallet |
Then come the rules that trip up new bots. All of them are in Hyperliquid’s API docs:
Do it from your main wallet, in the Hyperliquid app or with an approveAgent action in code. Give it a name and an expiry.
The bot signs with the API wallet’s private key and reads data with your account’s address, since the API wallet itself holds nothing.
It pulls prices, the order book and your positions from the info endpoint, or streams them over the websocket.
The order goes to the exchange endpoint and lands in your account, as if you had placed it by hand.
The API wallet can trade your balance but can’t withdraw it. To stop the bot, cancel its orders, then replace the API wallet or let it expire.
You can rehearse all of this on Hyperliquid’s testnet, a copy of the exchange that runs on play money. Its faucet gives 1,000 mock USDC to addresses that have deposited on the real exchange.
A bot is a rule that runs without you. The rule decides what it is good at and how it loses money. Our fuller definition of a trading bot covers the wider taxonomy. These are the seven kinds you will meet.
| Bot | What it does | Works best when | How it fails |
|---|---|---|---|
| Grid | Places a ladder of buy orders under the price and sell orders over it, and re-places each order after it fills | Price swings back and forth inside a range | The price leaves the range. In a fall it keeps buying into a bigger loss; in a rally it sells out early. With leverage, a long fall can liquidate it. |
| DCA | DCA stands for dollar-cost averaging: it buys a fixed amount on a schedule or after each drop, sometimes adding bigger safety orders lower down | You want to build a position over time without timing it | A long decline: every safety order adds size at a worse price. On perps you also pay funding for as long as you hold. |
| Trend or signal | Buys or sells when a rule fires: a moving-average cross, a breakout, an RSI (relative strength index) level | The market trends | A choppy market: false breakouts and whipsaws cost a little each time, and fees and slippage add up. |
| TradingView webhook | Turns a TradingView alert into an order: TradingView sends a message to the bot’s web address when your alert fires | You already run your strategy on TradingView charts | Late, missed or doubled alerts, and scripts that repaint, which means their past signals change after the fact and look better than they trade. |
| Copy | Watches another wallet and repeats its trades in your account, scaled to your size | You have checked the trader’s record | You fill after the leader at a worse price, you take on the leader’s leverage, and today’s top leaders include some who were simply lucky. |
| Market making | Quotes both sides of the order book to earn the gap between buyers and sellers | You have fast infrastructure, capital and code you trust | It gets filled right before the price runs, then holds the wrong side of a trend. |
| AI agent | A language model reads data, decides and calls trading tools, often from plain-English instructions | You want analysis, tests and orders in one conversation | It can misread an instruction or act on a wrong read with confidence, so it needs hard limits and a record of every action. |
Two notes on the table. TradingView only sends webhook alerts from accounts with two-factor authentication turned on, and it cancels a request if the receiving server takes longer than three seconds, so a slow bot can miss a signal. And copy bots have their own guide: how copy trading on Hyperliquid works, from vaults to wallet mirrors.
Most of the choice comes down to one question: is the market ranging or trending? A grid wants a range and a trend bot wants a trend. The bot that did well last month can bleed the next one when the market changes character.
Hyperliquid has order types built in that cover some of the jobs people buy bots for:
What still needs a bot: a grid that re-arms itself, a DCA plan that runs for months, or any rule that watches indicators and acts while you sleep.
Once you know the rule, the next choice is who runs it. There are four ways, and the biggest difference between them is who holds which key.
| Hosted platform | Open source (GitHub) | Telegram bot | AI agent | |
|---|---|---|---|---|
| Custody | Usually your balance stays in your Hyperliquid account and the platform holds an API wallet key you approved | You hold every key, on a machine you run | Often a wallet the bot creates for you; who holds its key varies, so read the docs before you deposit | Depends on the product: an API wallet, a wallet the app creates for you, or none at all |
| Setup | Minutes: connect, pick a template, set limits | Hours to days: install, configure, keep a server running | Minutes, inside a chat | Minutes: you describe what you want |
| Control | The templates and settings it offers | Everything: you can read and change every line | The commands it offers | Wide, as long as it reads you right |
| Cost | A monthly plan, a fee on each trade, or both | Free software, plus a server and your time | Usually a fee on each trade | A plan or usage credits, sometimes a fee on each trade |
| Main risk | An outage, or templates that don’t fit your market | Your own bug, or a server that stops | Someone else holding your key, and fake copies of the bot | A wrong read, acted on fast |
Many apps charge through builder codes, a fee Hyperliquid adds to your fills and pays to the app. Before any app can charge one, you approve its maximum rate from your main wallet, and Hyperliquid caps it at 0.1% on perps and 1% on spot. You can revoke a builder’s approval at any time.
Hosted platforms run the bot on their servers and give you templates. On Hyperliquid, most offer grid, DCA and signal bots on perps, and many accept TradingView alerts. WunderTrading, for example, runs signal, grid and DCA bots and, on Hyperliquid, charges a builder fee of 0.035% per trade instead of a plan. Gainium runs DCA, grid and combo bots, free with a builder fee or on a paid plan without one; its help page notes that Hyperliquid doesn’t allow builder fees on sub-account or vault connections, so those need a plan. 3Commas, Bitsgap, Coinrule and Cornix connect to Hyperliquid too, each with its own mix of bots and plans (we compare that kind of bot with Ethy in Ethy vs trading bots). They connect in one of two ways: you create an API wallet and paste its key (Bitsgap, Cornix), or you connect your main wallet and sign an approval that lets the platform trade through the API (3Commas, Coinrule). Check, too, that the tool still runs: the hosted goodcryptoX service, which some 2026 lists still recommend, has closed.
If you code, start from what Hyperliquid itself points to: its official Python SDK, community SDKs in TypeScript and Rust, and CCXT, a library that speaks to many exchanges in one format. The Python SDK’s examples show how to approve an agent, place and cancel orders and trade for a vault or sub-account. If you’d rather start from a full bot, the big open-source frameworks support Hyperliquid: Hummingbot (spot and perp connectors with testnet, and grid, DCA, TWAP and market-making strategies), Freqtrade (your own strategy, with backtests and dry runs), Passivbot (a grid-style market maker for perps) and NautilusTrader. Read each one’s Hyperliquid notes. Freqtrade’s warn that Hyperliquid “only offers 5000 historic candles,” which limits backtests, and that market orders are simulated with limit orders up to 5% away. They also put the key rule plainly: “Don’t use your actual wallet private key for trading.”
Be careful with the rest of the results. On November 14, 2026, one of the most-starred repos in GitHub’s search for “hyperliquid trading bot” was three days old and already had 289 stars and 67 forks. Its code loads a hidden file into memory on Windows, with the names of the system functions it uses disguised as bytes, the way malware loaders work. Stars can be faked. Read the code, never run a downloaded program or “loader”, use an API wallet, and try anything new on testnet first.
Telegram bots trade with chat commands, and most create the wallet for you. PvP.trade trades Hyperliquid perps from Telegram groups, with wallet tracking and copy trading, and says private keys “are stored and encrypted by Turnkey,” a wallet-infrastructure company; you can export yours. Based pairs a Telegram bot with a trading app and says “We don’t store your keys.” Some are not for perps at all: HypurrFun is for launching and trading memecoins. Scammers clone popular bots, so open one only from the link on its official site.
An AI agent puts a language model in front of the bot. You say what you want in plain words; it reads the market, proposes or places trades and reports back. Senpi, for example, turns a described strategy into a running bot on Hyperliquid. It creates a wallet for each strategy that “signs inside Privy’s secure enclave, under rules that only allow trades and moves between your own Senpi wallets,” and you can export your wallet. Plain words make a rule easier to write and harder to pin down, so the checklist below matters even more. Our explainer on what an AI trading agent is goes deeper, and Ethy’s approach closes this guide.
For named products side by side, see our list of the best Hyperliquid trading bots. This guide stays on how they work.
A bot’s orders move the price less where more people trade. On November 14, 2026, Hyperliquid’s main exchange listed 179 active perps. The five biggest took 91% of the day’s volume, 105 of the 179 traded less than $1 million that day, and 145 capped leverage at 5x or lower.
| Market | 24h volume | Open interest | Max leverage |
|---|---|---|---|
| BTC | $4.71B | $2.42B | 40x |
| ETH | $1.37B | $1.22B | 25x |
| HYPE | $959M | $1.30B | 10x |
| SOL | $461M | $293M | 20x |
| ZEC | $419M | $315M | 10x |
| XRP | $119M | $81M | 20x |
| NEAR | $83M | $106M | 10x |
| PUMP | $53M | $38M | 10x |
| LIT | $48M | $88M | 5x |
| XMR | $43M | $96M | 5x |
Open interest is the value of all positions still open. The numbers come from Hyperliquid’s public API on November 14, 2026, at 20:52 UTC, and they move every day. Beyond the main exchange, HIP-3 exchanges (more on them below) listed 171 active markets, and trade.xyz had 98% of their volume. Its busiest were the XYZ100 index ($221 million), the S&P 500 ($212 million) and Nvidia ($151 million).
The lesson for a bot: a grid that is fine on BTC can push the price around on a market that trades a few million dollars a day. Size each order against the order book, not against your balance.
Leverage lets you hold a position bigger than the money behind it, the margin: at 10x, $1,000 of margin holds a $10,000 position. Liquidation is the forced closing of that position when the margin can no longer cover the loss. On Hyperliquid it happens when your account value falls below the maintenance margin, which is half the initial margin at the market’s max leverage: 1.25% of the position on a 40x market such as BTC, up to 16.7% on a 3x market. If the order book can’t close the position and your equity drops below two-thirds of that margin, a backstop vault takes the position over and the margin left is lost. Leverage is checked only when you open. After that, the docs say, “the user is responsible for monitoring the leverage usage to avoid liquidation.” A bot doesn’t change that.
| Leverage | BTC drop that liquidates an isolated long |
|---|---|
| 3x | about 32.5% |
| 5x | about 19.0% |
| 10x | about 8.9% |
| 20x | about 3.8% |
| 40x | about 1.3% |
An isolated position has its own margin, apart from the rest of your account. The table uses Hyperliquid’s liquidation formula and BTC’s 40x cap, and it leaves out fees, funding and the higher margin that very large positions need. Grids and DCA bots add to losing positions by design, so their liquidation price creeps closer as they buy. Give every bot a maximum position size and a stop.
Use an API wallet, never your main private key or your seed phrase, the list of words that restores your wallet. Hyperliquid’s security advice is to never share either one and never type them into a website. Give each bot its own API wallet, with an expiry. And remember what an API wallet can still do: it can’t withdraw, but it can trade. A stolen trading key can open a large position in a thin market at an awful price, with someone else on the other side of the trade.
Slippage is the gap between the price you expect and the price you get. Through the API, a market order is an aggressive limit order that fills at once or cancels, and Hyperliquid’s Python SDK prices it 5% from the current price by default, so set a tighter cap in your bot. A market stop or take profit on Hyperliquid accepts up to 10% slippage. A limit stop protects the price but may not fill at all in a fast move: Hyperliquid’s own example is a stop at $10 that sits unfilled when the price jumps from $11 straight to $9.
Funding is the payment between longs and shorts that keeps a perp’s price near the spot price. On Hyperliquid it is paid every hour. Its base rate has longs paying shorts 0.00125% an hour, which the docs put at 11.6% a year, and the real rate moves with the gap between the perp and spot, capped at 0.5% an hour. A DCA bot that holds a long perp for months pays or earns funding the whole time. A spot purchase doesn’t.
Many bot losses come from plain bugs, not the market: an order sent twice after a restart, two processes fighting over one nonce, a websocket that dropped without anyone noticing, or a rate limit that leaves the account one action every 10 seconds in the middle of a fast move. Put expiry times on orders, set the dead man’s switch, log every action, and run new code on testnet first.
If a losing account goes below zero in an extreme move, Hyperliquid closes positions on the winning side, ranked by profit and leverage, at the previous mark price. This is auto-deleveraging, or ADL. It keeps the exchange solvent, and it can close a bot’s best trade early, so the code should expect positions to close without an order of its own.
HIP-3 lets builders deploy their own perp exchanges on Hyperliquid, which is how stocks, indices and commodities trade there. On a HIP-3 market, the deployer defines and runs the oracle, the price feed the contract tracks, sets the leverage limits and can settle the market at the mark price. On trade.xyz, the busiest, the oracle follows outside prices while the underlying market is open. When that market is closed, the oracle moves with trading on trade.xyz itself, within set bounds, and when outside prices return it reverts to them “on the next tick”. A stock that opens far from Friday’s close can jump past your bot’s stop in one step. The bounds have a second effect: in trade.xyz’s words, “if a trader’s liquidation price lies outside the active price bounds, their position cannot be liquidated while those bounds are in effect,” so the risk waits for the reopen. Our guide on how to trade stocks on Hyperliquid covers these markets in full.
Ask these before you connect anything. A good tool answers each one in writing.
Ethy is an AI trading agent: you ask it in plain English, before you trade and while a position is open. It reads a chart on four timeframes (daily, 4-hour, 1-hour and 15-minute), finds out why a market moved with a web search that cites its sources, tests a rule on up to 365 days of past data, and knows your trading plan, the rules you set for yourself, which it reads first in every chat.
It can also do the trading side, on Hyperliquid’s main exchange for crypto and on trade.xyz for stocks, commodities, indices and FX:
What it doesn’t do: copy another wallet, take TradingView alerts, or trade HIP-3 exchanges other than trade.xyz. And because Hyperliquid nets all trades in one market into one position, Ethy runs one automation per market per wallet.
The one whose rule fits the market you trade and whose key model you trust. Developers often start from Hyperliquid’s official Python SDK; traders who don’t code use a hosted platform, a Telegram bot or an AI agent. Our list of Hyperliquid trading bots compares named products. Whatever you pick, run the checklist in this guide before you connect it.
Not if it only holds an API wallet. Hyperliquid’s official SDK says an API wallet does not have permission to transfer or withdraw funds. It can still trade your balance, losses included, so give it limits. A bot that holds your main private key or seed phrase can do anything with the account, withdrawals included.
Yes, from hosted platforms and open-source frameworks. On the exchange itself, a scale order places a ladder of limit orders once and a TWAP order trades in slices over up to 7 days, but neither re-arms like a grid or keeps buying for months like DCA.
Yes, through a bot that accepts TradingView webhooks. TradingView sends an HTTP message to the bot’s web address when your alert fires, and the bot places the order with your API wallet. TradingView requires two-factor authentication for webhook alerts and cancels a request that takes the bot more than three seconds to handle.
No. Hosted platforms, Telegram bots and AI agents need no code. Coding gives you full control: Hyperliquid’s Python SDK and the community TypeScript and Rust SDKs cover the whole API, and the CCXT library supports Hyperliquid too. If you code, start on testnet with an API wallet.
Hyperliquid recovered a different signer from your signature than the one you meant. Its signing docs list the usual causes: mixing up its two signing schemes, fields in the wrong order, trailing zeros in numbers, and upper-case letters in addresses. Use an SDK rather than signing by hand, and check that the API wallet is approved and hasn’t expired.
Usually because it asks for the API wallet’s balance. An API wallet only signs, and Hyperliquid’s docs warn that querying with its address leads to an empty result, so pass your main account’s address. In unified or portfolio-margin mode, balances also show in the spot account rather than in each perp account, which trips up older bot code.
A bot is as good as its rule, its costs and its limits. It removes hesitation and missed signals, but it also repeats a mistake every time the rule fires. Test the rule on past data, run it small or on paper first, and judge it by its worst drawdown as well as its return.
Trading involves risk, including the loss of your money. Ethy AI is a research and trading tool and does not give investment advice or recommendations. Backtests are simulations on historical data and do not predict future results.
Deploy your AI trading agent in one click.
