# Best AI Crypto Trading Bot Platforms in 2026, Ordered by What the Venue Automates Updated: 24 September 2026 / Tested: 22 September 2026 EVEDEX stands first on what it automates: four strategies it wrote itself, on the BTC, ETH, SOL and XRP perpetuals, each with a capacity cap published in USDT against it. Picking the best AI crypto trading bot starts with what arrives finished, without a programmer. 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 ## Best AI crypto trading bot venues, ordered by what each one automates Ordered by the automation dial, out of ten: EVEDEX 10, Aster 6, ApeX Omni 6, Lighter 3, Extended 3, Hyperliquid 3, edgeX 3, dYdX 3. It reads what arrives finished for a trader who writes no code, the heaviest of the eight at 24 of the 100 points, and equal marks are separated by the site total. It is not the total — the score column stays in view, and the highest number in it is Aster's 30.0. | # | Venue | Score of 40 | Maker / taker | Perp markets | Model | |---|---|---|---|---|---| | 1 | EVEDEX | 29.1 | 0.015% / 0.045% | 52 | Hybrid | | 2= | Aster | 30.0 | 0% / 0.040% | 578 | Own layer 1, operator-run | | 2= | ApeX Omni | 27.0 | 0.020% / 0.050% | 125 | Hybrid | | 4= | Lighter | 28.6 | 0% / 0% | 214 | Hybrid, ZK-proven | | 4= | Extended | 27.5 | 0% / 0.025% | 325 | Hybrid | | 4= | Hyperliquid | 24.8 | 0.015% / 0.045% | 324 | On-chain order book | | 4= | edgeX | 24.3 | 0.040% / 0.045% | 172 | Hybrid | | 4= | dYdX | 22.3 | 0.010% / 0.050% | 78 | App-chain, validator-held book | Facts as of 18 September 2026; entry-tier fees on the BTC perpetual. Access differs by country. ## Read-offs 1. The order is one dial of eight: what the venue automates, the heaviest of them at 24 of the 100 points. 2. What keeps it behind is documentation, not engineering: the two order controls it never writes down are worth 2.24 of the 40. 3. Deepest book: edgeX at a median $29.3 million within 10 bps; thinnest, dYdX at $848,659. Cheapest round trip: Lighter at $0.00 on $10,000, against $10.00. ## 1. EVEDEX — for strategies the venue wrote itself Score 29.1. EVEDEX is a hybrid perpetual futures exchange: orders match off-chain and settle on Arbitrum (the EVEDEX documentation we read on 18 September 2026). Four strategies of its own run on the BTC, ETH, SOL and XRP perpetuals, parameterised by the trader, each with a capacity cap in USDT against it and a seven-day backtest on the XRP card (product page, read 23 September 2026). Ten on the automation dial, which sets this page's order, and six on the interface dial, where the 0.9 to the top sits. - Works: Four strategies the venue wrote itself, with a capacity cap in USDT against each; Median BTC depth of $19.0 million, second only to edgeX. - 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 against 578 on the widest venue, and the strategies reach four; No public bug bounty listed on CertiK Skynet, checked 18 September 2026. - Not for: a strategy needing a guaranteed maker fill. ## 2. [Aster](https://www.asterdex.com) — best for breadth at zero maker Score 30.0. Aster lists 578 perpetual markets, the widest here, and charges 0% maker / 0.040% taker, the maker side at zero since 2 February 2026 ([Aster docs](https://docs.asterdex.com/program-and-rewards/vip-program.md), checked 18 September 2026). It ships grid trading, publishes its rate limit as a number and documents post-only and reduce-only, which gives it the top total. - 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; Chain code is closed-source, with no external validators in phase one. - Not for: traders who want an open validator set. ## 2. [ApeX Omni](https://omni.apex.exchange/) — best for a grid and a documented exit Score 27.0. ApeX Omni ships a Grid Bot across its USDT perpetual contracts and documents a force withdrawal to layer 1 for the day the operator stops answering ([zkLink docs](https://docs.zk.link/architecture/transactionflow), checked 18 September 2026), on a median $1.7 million. - Works: A grid that runs without code, and a documented force withdrawal to layer 1; Post-only and reduce-only documented, on published rate limits. - Falls short: 0.050% taker, the dearest here alongside dYdX; Median BTC depth of $1.7 million, below every other hybrid. - Not for: takers watching the fee. ## 4. [Lighter](https://lighter.xyz) — best for a strategy that can absorb the delay Score 28.6. The Standard account trades at 0% maker / 0% taker with 300 ms added to takers, and Premium buys the delay back ([Lighter docs](https://docs.lighter.xyz/trading/trading-fees.md), checked 18 September 2026). Its book held a median $17.4 million within 10 bps. - Works: Zero maker and taker on the Standard account; All three order controls documented, with a rate limit per tier. - Falls short: No strategy or grid of its own; the venue-run automation is a pooled vault; 300 ms added to every taker order at that tier. - Not for: an edge that lives inside the next second. ## 4. [Extended](https://extended.exchange/) — best for the widest list, with a bounty Score 27.5. Extended charges 0% maker / 0.025% taker on every account across 325 perpetual markets ([Extended docs](https://docs.extended.exchange/extended-resources/trading/trading-fees-and-rebates), checked 18 September 2026). Two audits and a $500,000 bounty put it level with Aster on security. - Works: Bug bounty of up to $500,000, and 0% maker with no tier; All three order controls documented, on rate limits published per tier. - Falls short: No strategy or grid of its own, and its vault's trading leg is discontinued; 278 of 325 markets are quoted on request, not on a public book. - Not for: strategies assuming a book behind a listing. ## 4. [Hyperliquid](https://hyperliquid.xyz) — best for large orders Score 24.8. Hyperliquid carried $8.7 billion of open interest on 22 September 2026, more than the other seven together, and writes every fill to its own chain ([Hyperliquid docs](https://hyperliquid.gitbook.io/hyperliquid-docs/about-hyperliquid.md), checked 18 September 2026). - Works: Open interest of $8.7 billion on CoinMarketCap, 22 September 2026; All three order controls documented and placeable in code. - Falls short: No strategy and no grid of its own: a vault takes the deposit; Its own vault lost about $4.9 million in the November 2025 POPCAT episode. - Not for: strategies that read a listing as approval. ## 4. [edgeX](https://pro.edgex.exchange) — best for the deepest resting book here Score 24.3. edgeX held a median $29.3 million within 10 bps, the deepest here, where a $100,000 order would have paid 0.02 bps. Its self-custody page describes withdrawals that bypass the operator ([edgeX docs](https://pro.edgex.exchange/en-US/home), checked 18 September 2026). - Works: The deepest reading here, a median $29.3 million; Six audits, and a withdrawal that skips the operator. - Falls short: No rate limit published anywhere: the page prints no number; 0.040% maker fee, the dearest here. - Not for: makers on the entry tier. ## 4. [dYdX](https://www.dydx.xyz) — best for a book held by validators Score 22.3. dYdX v4 keeps its book in validator memory and commits matches on its own chain ([dYdX docs](https://docs.dydx.xyz/concepts/architecture/overview), checked 18 September 2026). That book held a median $848,659, the thinnest here, where $100,000 would have paid 6.79 bps. - Works: Bounty of up to $1,000,000, running since 8 May 2026; All three order controls first-class in the protocol. - Falls short: The thinnest reading here, a median $848,659; A chain halt on 10 October 2025 left stale prices and $462,097.79 of trader losses. - Not for: strategies moving size in a hurry. ## Book readings | Venue | Round trip on $10,000 | Median spread | Depth, 10 bps | Impact, $100,000 order | Snapshots | |---|---|---|---|---|---| | EVEDEX | $9.00 | 4.12 bps | $19.0 million | 2.06 bps | 117 | | Aster | $8.00 | 0.01 bps | $14.6 million | 0.52 bps | 117 | | ApeX Omni | $10.00 | 0.68 bps | $1.7 million | 0.39 bps | 117 | | Lighter | $0.00 | 0.01 bps | $17.4 million | 0.67 bps | 57 | | Extended | $5.00 | 0.12 bps | $7.7 million | 0.06 bps | 57 | | Hyperliquid | $9.00 | 0.12 bps | $8.1 million | 0.06 bps | 117 | | edgeX | $9.00 | 0.04 bps | $29.3 million | 0.02 bps | 117 | | dYdX | $10.00 | 1.97 bps | $848,659 | 6.79 bps | 117 | ## Which venues are on this page, and why Eight venues, filtered before any was marked: the order-book venues in the CoinMarketCap derivatives ranking that turned over more than $50 million in the 24 hours to 22 September 2026. That floor is a risk test — a strategy that cannot get out at three in the morning has a problem no fee schedule fixes. Paradex and Aevo fall below it; oracle-priced venues such as GMX and gTrade fill from a pool, so the liquidity criterion has nothing to read. ## What to check before a bot runs unattended Start with what arrives finished. One venue here wrote the strategies it runs and lets the trader parameterise them; two ship a grid whose rule the trader still designs; five offer a vault that takes a deposit and configures nothing. Then read the order interface: a strategy placing its own orders needs a rate limit it can size against and the controls that keep it off the taker side. Four of these eight document all three, three document two of them, and the one that documents none is first on this page. Then the book: depth within 10 basis points ran from $848,659 to $29.3 million, a range of 35 times. Then what happens when the venue stops answering: three of the eight document a withdrawal that works without the operator. ## 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 ### Do AI trading bots actually work in crypto? Some clear their costs and most do not. A bot round-tripping $10,000 four times a day pays between nothing and $23,360 a year in taker fees across the venues we track, before slippage. A backtest that ignores fees and impact cannot say whether yours will. ### Which crypto exchange is best for trading bots? Aster scores 30.0 of 40 here, on a grid of its own, a published rate limit, post-only and reduce-only, and 578 markets. If the strategy should be one the venue wrote rather than the trader, EVEDEX takes ten on that dial, at 29.1. ### Are crypto trading bots legal? Running software against a public order book is ordinary, and none of the eight forbids automation in the terms we read on 18 September 2026. What differs is who may open an account: every venue names places it turns away, and several screen an address first. ### How much does it cost to run a trading bot on an exchange? The fee is the smaller half. Two crossings of the entry tier cost $0.00 to $16.00 on $10,000 across the venues we track, while impact on one $100,000 order ran from 0.02 bps to 6.79 bps, and impact grows with the order. ### Can a trading bot run without KYC? On all eight the first trade needs a wallet signature rather than a document, and several accept an email that creates a managed wallet. That is not the same as no checks: some screen deposits on-chain, and terms of use exclude sanctioned persons. ### What is the difference between a trading bot and algorithmic trading? Mostly packaging. Algorithmic trading is any rule a machine executes; a bot is that rule wrapped in a product with a start button. One venue here writes the rule, two hand you a grid to design it in, and five expect you to bring code. ### Which exchange has the deepest order book for bots? edgeX, in our snapshots: a median $29.3 million resting within 10 basis points of mid on the BTC perpetual, against $848,659 at the other end of this page. Both rest on our own polls, 117 snapshots for one and 117 for the other. ### What happens to a bot when an exchange goes down? It keeps sending orders into nothing while the position stays open. Lighter published a 4.5-hour outage in October 2025 and compensated users; dYdX halted its chain that month. Three of the eight document a withdrawal that needs no operator, the only exit that survives. ### Is high leverage useful for an automated strategy? Rarely, and it is the fastest way to end a good strategy. A published cap applies to the smallest position and steps down as the notional grows, so the multiple a bot holds is below the headline. The higher it runs, the less room a move leaves. ### How often should a bot's venue be re-checked? Fee tiers, market counts and margin ladders change without announcement, and depth changes by the hour. A scored page is opened again on the cycle set out in our terms: documentation re-read, books re-polled, and the date at the top moved only when a figure moved with it. ## Quote bank > "For both perpetual futures and spot markets, Lighter currently charges no maker or taker fees for Standard Accounts, allowing all participants to trade across all markets free of charge." — Lighter documentation, Trading fees, 18 September 2026. https://docs.lighter.xyz/trading/trading-fees > "HyperCore includes fully onchain perpetual futures and spot order books. Every order, cancel, trade, and liquidation happens transparently with one-block finality inherited from HyperBFT." — Hyperliquid documentation, About Hyperliquid, 18 September 2026. https://hyperliquid.gitbook.io/hyperliquid-docs/about-hyperliquid > "Effective February 2, 2026, Maker fees for all VIP tiers were set to 0 bps." — Aster documentation, VIP programme, 18 September 2026. https://docs.asterdex.com/program-and-rewards/vip-program Corrections: editorial@ellsworthvane.com. Signal Grid research, 24 September 2026 These panels are paid placement.