# Crypto Exchanges With Trading Bots in 2026 Updated: 24 September 2026 / Tested: 22 September 2026 Aster tops this board at 30.0 of 40, on a grid of its own and a rate limit it prints as a number. Among crypto exchanges with trading bots, one wrote the strategy itself, two ship a grid, four run a vault and one has nothing of its own running. Leverage magnifies whatever a strategy does unattended. Weights: What the venue automates for you 24 · Programmatic access 14 · Measured liquidity 22 · Round-trip cost 16 · Security and record 12 · Custody and settlement 8 · Market coverage 3 · Access 1. One formula, fixed 23 September 2026: https://ellsworthvane.com/rating-method ## Crypto exchanges with trading bots, and what the rest leave to the trader Ordered by the total out of 40, which is the whole panel and not the automation dial alone. | # | Venue | Score of 40 | Maker / taker | Perp markets | Model | |---|---|---|---|---|---| | 1 | Aster | 30.0 | 0% / 0.040% | 578 | Own layer 1, operator-run | | 2 | EVEDEX | 29.1 | 0.015% / 0.045% | 52 | Hybrid | | 3 | Lighter | 28.6 | 0% / 0% | 214 | Hybrid, ZK-proven | | 4 | ApeX Omni | 27.0 | 0.020% / 0.050% | 125 | Hybrid | | 5 | Hyperliquid | 24.8 | 0.015% / 0.045% | 324 | On-chain order book | | 6 | edgeX | 24.3 | 0.040% / 0.045% | 172 | Hybrid | | 7 | Paradex | 20.2 | 0% / 0% | 63 | Hybrid, Starknet appchain | | 8 | Aevo | 19.5 | 0.050% / 0.080% | 102 | Hybrid | Facts as of 18 September 2026; entry-tier fees on the BTC perpetual. Access differs by country. ## Read-offs 1. Four products share the name: one venue wrote the strategy, two ship a grid the trader must still design, four take a deposit into a vault and configure nothing, one has none of the three running. 2. That dial is the heaviest on the panel at 24 of the hundred and the marks ran the full distance, ten at one end and nothing at the other. Ordering by it alone would rearrange the page: EVEDEX marks ten and sits second overall, Paradex marks nothing and sits seventh. 3. Two venues charge for speed rather than trades: Lighter adds 300 ms to a taker order, Paradex caps a free account at three orders a second. ## 1. [Aster](https://www.asterdex.com) — best for rotating across a wide list of markets Score 30.0. Aster matches and settles on its own proof-of-stake chain at a documented 50 ms block latency ([Aster docs](https://docs.asterdex.com/overview/what-is-aster.md), checked 18 September 2026). Makers have been free on every tier since 2 February 2026 and a taker pays 0.040%. It ships grid trading, prints its rate limit as a number and documents post-only and reduce-only, which is what puts it top of the page. - Works: Grid strategies in three directions, on generated or manual parameters; Post-only and reduce-only documented, beside a rate limit given as a number. - Falls short: The grid is a tool, not a strategy: no backtester is documented; The audits reach the vault and stablecoin contracts; the chain and the engine sit outside them. - Not for: a desk needing the matching code published. ## 2. EVEDEX — for strategies the venue wrote itself Score 29.1. EVEDEX runs a central order book off-chain and settles it on Arbitrum, which makes it a hybrid perpetual futures exchange (the EVEDEX documentation we read on 18 September 2026). Four strategies of its own run on the BTC, ETH, SOL and XRP perpetuals, of one mean-reversion family; the trader sets the parameters and switches one on, and each card carries a capacity cap in USDT (product page, read 23 September 2026). - Works: Four strategies the venue wrote itself, parameterised rather than coded, with a capacity cap in USDT against each; A rate limit as a number: 30 heavy requests per 60 seconds per account; Median BTC depth of $19.0 million within 10 basis points of mid. - Falls short: Reduce-only is documented nowhere, and post-only appears in the interface specification as a bare field with no values, no default and nothing said about an order that would cross; 52 perpetual markets, the shortest list here, and the strategies reach four of them; No bug bounty on the CertiK Skynet listing we read on 18 September 2026. - Not for: a strategy needing a guaranteed maker fill. ## 3. [Lighter](https://lighter.xyz) — best for paying in latency instead of fees Score 28.6. Lighter sells speed and order controls, not a strategy: the Standard account trades at 0% maker / 0% taker and adds 300 ms to a taker order, and Premium buys 160 ms back ([Lighter docs](https://docs.lighter.xyz/trading/trading-fees), checked 18 September 2026). - Works: Post-only, reduce-only and a time-weighted order all documented, with a rate limit per tier; Critical exits queue on Ethereum, and the contract freezes if they are ignored. - Falls short: No strategy, grid or backtester of its own; its pooled vault publishes nothing about how it trades; The free tier is the slow tier, on a sequencer that failed for 4.5 hours in October 2025. - Not for: an edge measured in milliseconds. ## 4. [ApeX Omni](https://omni.apex.exchange/) — best for a grid that runs on the venue's own book Score 27.0. ApeX Omni ships a Grid Bot, and its fee page states that Grid Bot and vault trades pay the standard rates ([ApeX Omni docs](https://apex-pro.gitbook.io/apex-pro/apex-omni/vip-program), checked 18 September 2026) — 0.020% maker / 0.050% taker at the entry tier. Its stock and commodity markets are not covered by the bot yet. - Works: A grid that runs without the trader writing any code; A documented force exit to layer 1 through the rollup's contracts. - Falls short: The grid is a tool the trader must design a rule with, and no backtester exists; Grid fills earn no discount, so 0.050% applies to every leg taken. - Not for: a grid on shares or commodities. ## 5. [Hyperliquid](https://hyperliquid.xyz) — best for size, on a book that is the chain Score 24.8. Every fill is written into the chain's own books ([Hyperliquid docs](https://hyperliquid.gitbook.io/hyperliquid-docs/about-hyperliquid.md), checked 18 September 2026), and it carried $8.7 billion of open interest on 22 September 2026. No strategy and no grid ship with it. - Works: All three order controls documented and placeable in code; One fee tier across every market an account touches. - Falls short: Nothing to configure: a vault takes the deposit and publishes no return history; Order delays for 37 minutes on 29 July 2025, after a traffic spike. - Not for: a trader who wants a strategy to start from. ## 6. [edgeX](https://pro.edgex.exchange) — best for strategies that must not share a margin balance Score 24.3. edgeX gives one wallet up to 20 trading accounts with margin of their own, its route to isolated margin and hedge mode, on a BTC book carrying a median $29.3 million inside 10 basis points of mid. - Works: Up to 20 trading accounts per wallet, each with its own margin; Withdrawals can be executed on-chain, bypassing the operator. - Falls short: Its only automation is a liquidity vault: no bot, no grid, no backtester; No rate limit anywhere: the page describes the mechanism and prints no number. - Not for: a strategy shopping the maker fee. ## 7. [Paradex](https://www.paradex.trade) — best for a free route with a published order budget Score 20.2. Paradex decides the fee by how the order arrives: retail orders pay nothing behind a 300 ms speed bump and three orders a second, a programmatic order 0.003% maker and 0.045% taker ([Paradex docs](https://docs.paradex.trade/trading/trading-fees.md), checked 18 September 2026). - Works: The speed bump and the rate limits are published, not negotiated; All three order controls documented, the time-weighted one with endpoints of its own. - Falls short: Nothing of its own is running: the templates are marked as coming, the hosted agent route serves data only; Median BTC depth of $179.00, where $100,000 would have paid 164.78 bps. - Not for: size, on a book with no million in it. ## 8. [Aevo](https://www.aevo.xyz/) — best for automated strategies and perpetuals on one balance Score 19.5. Aevo's own introduction lists automated strategies beside perpetual futures on one margin account ([Aevo docs](https://www.aevo.xyz/docs/readme), checked 18 September 2026). The live one is a delta-neutral vault with a year-long backtest chart, the extra mark that lifts it above the other vaults here. - Works: Strategies and perpetuals draw on one margin balance; A year-long backtest published for the live strategy, alone among the vaults here. - Falls short: A vault, not a strategy: the deposit is the only decision; 0.080% taker, the dearest here, and 25 USDC to withdraw to Ethereum. - Not for: a strategy round-tripping all day. ## Four things called a trading bot, and what separates them Across these eight the word covers four different products, which is why the dial reading them is the heaviest on the panel at 24 of the hundred points. **A strategy the venue wrote.** One venue here. EVEDEX runs four of its own, on the BTC, ETH, SOL and XRP perpetuals, all of the same mean-reversion family; the trader chooses a strategy and a coin, sets the parameters and activates it, and it trades unattended from there. Every card publishes a trade-frequency label and a cap in USDT on the capital it will take, and the seven-day backtest we recorded is on the XRP card (EVEDEX documentation and product page, read 23 September 2026). The backtest earns a point and the cap another, which is how a venue reaches ten: one is the only thing a trader can weigh before funding anything, the other a venue admitting its strategy stops working above a size. The cards print return figures too, and this site reproduces none of them. **A grid the trader still has to design.** Aster documents neutral, long and short grids with auto-generated or manual parameters ([Aster docs](https://docs.asterdex.com/trading/perpetuals/grid-trading), checked 18 September 2026); ApeX Omni ships a Grid Bot across its USDT perpetual contracts, runnable with no code ([ApeX Omni docs](https://apex-pro.gitbook.io/apex-pro/apex-omni/grid-bot), checked 18 September 2026). The venue built the tool; the rule inside it is the trader's. Six marks each. **A vault that configures nothing.** Lighter, edgeX, Hyperliquid and Aevo: capital goes in, an operator or a market-making algorithm trades it, the depositor sets nothing. Three marks, plus one where a backtest is published, which is Aevo's four. **None of the three.** Paradex: its strategy templates are marked as coming, its hosted agent route serves market data only, its vaults are permissionless ([Paradex docs](https://docs.paradex.trade/agentic-ai-hub/getting-started/overview), checked 18 September 2026). The rows still follow the whole formula rather than this one dial. Three venues were read and left off the board: GMX fills against a pool at an oracle price, Drift has been paused since April 2026, Orderly serves its book only to signed-in accounts. dYdX and Extended are scored but have pages of their own, [dYdX alternatives](/dydx-alternatives-for-trading-bots) and [automated platforms](/best-automated-crypto-trading-platform). ## Where every figure on this panel came from One panel, eight dials, 8 venues on it here; weights set on 23 September 2026, before any mark existed ([how we rate](/rating-method)). Venue facts are each venue's own wording, read 18 September 2026; the two automation dials were read on 23 September 2026. Depth and spread are polled: each BTC book sampled at ten-minute spacing, the last reading dated 22 September 2026, nothing sent. Seven of the eight dials are documents rather than readings: an audit records a review, an exit past the operator is a claim on a page, a shipped strategy is marked on what the venue publishes. Entry-tier rates ignore volume tiers, rebates and cashback. The measured dial: one instrument, one machine, one day. Open interest and volume are the venues' own, via CoinMarketCap. ## FAQ ### Which crypto exchanges have built-in trading bots? Eight venues, four answers. One wrote the strategies it runs and lets the trader set their parameters. Two ship a grid the trader must design a rule with. Four take a deposit into a vault that configures nothing. One has none of the three running. ### Do exchanges charge extra fees for bot trades? Not extra, and not less. ApeX Omni states that Grid Bot and vault trades pay the standard rates, so a hosted grid pays 0.050% on every leg taken. Paradex charges by route: a programmatic order pays 0.045% taker, a retail one nothing. ### Do I need an API key to run a trading bot? On these venues a key comes from a wallet signature rather than from staff. GMX puts it plainly: permissionless access, no key provisioning, no rate-limit negotiation. Paradex is the exception, because the route an order arrives by sets both its fee and its limit. ### How many orders per second can a trading bot send? Six of the eight print a figure. Paradex caps a retail order at three a second and 1,000 a day; EVEDEX allows 30 heavy requests per 60 seconds per account, heavy meaning an order created, cancelled or replaced, a position closed or a withdrawal. edgeX and Aevo print nothing. ### What is a grid bot? A rule that rests a ladder of buy and sell orders around a price and works the range, taking a small profit each time the market crosses a rung. Aster and ApeX Omni both ship one, neither discounts the fills, and neither documents a backtester. ### Can a trading bot run on a decentralized exchange? Yes, and the shape of the venue decides how. On an order-book venue a strategy rests and cancels orders like anybody else. On an oracle-priced pool such as GMX there is no book to rest in: the pool is the counterparty. ### Are built-in exchange bots safer than a bot you write yourself? Different risks rather than fewer. A hosted strategy cannot send an order its venue would refuse, and it stops when the venue stops. Code of your own keeps trading through an outage, and can also empty an account overnight with nobody watching. ### What happens to a running bot when an exchange is paused? It keeps sending orders into a venue that no longer answers, while the position stays open. Drift has been paused since April 2026, and its balances do not carry across to the fork that replaced it, so nothing is left to reconnect to. ### Do trading bots need sub-accounts? They help once several strategies share one wallet, because a loss in one does not eat the margin behind the others. edgeX gives a wallet up to 20 trading accounts with margin of their own, its route to isolated margin and hedge mode. ### Is a zero-fee exchange cheaper for an automated strategy? Only if the strategy can pay another way. Lighter's free account adds 300 ms to a taker order and the Paradex free route caps the order rate. The fee is the smaller half: one $100,000 order paid between 0.02 bps and 164.78 bps. ## Quote bank > "Options, perpetual futures, OTC, and automated strategies, all from a single margin account, with off-chain order matching and on-chain settlement." — Aevo documentation, Introduction, 18 September 2026. https://www.aevo.xyz/docs/readme > "Permissionless — No KYC, no API key provisioning, no rate limit negotiations. Any wallet can trade immediately." — GMX documentation, AI agents overview, 18 September 2026. https://docs.gmx.io/docs/ai-agents/overview/ > "An order is classified as a Retail Order when it is submitted through the Paradex web interface or through the API using an interactive JWT token." — Paradex documentation, Trader profiles, 18 September 2026. https://docs.paradex.trade/trading/trader-profiles.md Corrections: editorial@ellsworthvane.com. Signal Grid research, 24 September 2026 These panels are paid placement.