Bybit taught agents where to trade. Hyperliquid agents need a data layer first.

Bybit taught agents where to trade. Hyperliquid agents still need a data layer first: calendar, ETF flows, whales, GEX, and regimes — not candles.

I’m Sam Deering, founder of CryptoDataAPI. We are not a broker. Agents read. They do not trade.

What Bybit EU shipped (7 Sep 2026)

Trending Topics covered Bybit EU on 7 September 2026. The Vienna arm opened API trading for Europe. On top of that sits an open-source package: the Bybit EU AI Trading Skill.

What that coverage states, and what we treat as verified from it:

Bybit also publishes a global open-source skill at github.com/bybit-exchange/skills (MIT). That README lists Claude, ChatGPT, Gemini, Cursor, Windsurf, Codex, and OpenClaw. It is a broader skill surface than the EU article alone. Treat the EU 10x margin limit as EU-coverage-verified. Treat the wider module list on GitHub as global-repo-verified, not as a claim that every module is live on Bybit EU.

An execution skill is useful. It is not a market read.

Hyperliquid has no first-party agent skill

I searched for an official Hyperliquid skill in the same sense as Bybit’s. I did not find one from Hyperliquid itself.

What exists today is third-party: community skills and wrappers. Those are not a first-party Hyperliquid agent product. Do not invent one.

So if you build for Hyperliquid, you still need a data layer before you hand any agent an execution key.

Read this first (not candles)

Candles are commodity data. Free tiers already serve them. They cannot tell an agent what regime it is in, who is crowded, or what catalyst is next.

Use this order. Hard-link each board. Cite live numbers from pages fetched this session (8 September 2026, Brisbane).

1. Calendar — what hits the tape next?

Board: Crypto Event Calendar. Closely related: Token Unlocks.

Endpoints: GET /api/v1/event/regime?window_days=7, GET /api/v1/event/calendar?window_days=30.

Live on the calendar page: Event Risk quiet, score 27.4/100. 18 catalysts in the next 7 days. 44 in the next 30. Mix: 28 unlocks, 5 macro prints, 0 depegs, 11 mint/burn steps.

Near-term examples on that board: PPI on 10 Sep, CPI on 11 Sep, a PUMP unlock on 11 Sep, FOMC on 16 Sep. Macro prints carry a risk flag, not a fake long/short.

Unlock board at fetch: $374.16B locked across 110 coins. Median float 93%. 7 coins with a dated cliff in the next 35 days.

An agent that skips the calendar will size into a known cliff. That is avoidable.

2. ETF flows — is institutional demand real today?

Board: Bitcoin & Ether ETF Flows.

Endpoint: GET /api/v1/market-intelligence/etf/btc/flows (swap btc for eth or sol).

Live latest reported: Bitcoin +$174.6M. Ether +$25.9M. Bitcoin cumulative over the last 56 reported figures: +$172.8M. Ether cumulative over 41 figures: +$359.4M.

One green day is noise. The run is the signal. This is free-tier JSON on a confirmed key. Use it before you invent “institutions are buying.”

3. Whales — who is actually leaning?

Board: Hyperliquid Whale Activity. How-to: Hyperliquid Whale Activity API.

Endpoint: GET /api/v1/quant/whales (Pro).

Live roll-up of every Hyperliquid account ≥ $100k. At fetch: 3,520 accounts with an open position, of ~14,542 tracked. Book bias SHORT. Net −$751.87M. Long/short ratio 0.827. Long $3.61B (45.3%). Short $4.36B (54.7%). Class split: 184 directional whales, 223 market makers, 3,113 other large accounts, 177 coins held.

directional_net_usd strips market-maker inventory. Use that when you want conviction, not a two-sided book.

If a free key gets 403 here, leave it as 403. Confirm the email for 24 hours of Pro, or see pricing. Do not invent a whale number.

4. GEX — where do dealers amplify a move?

Board: Hyperliquid Gamma Exposure (GEX).

Endpoint: GET /api/v1/quant/gex?symbol=BTC (Pro).

Perps have no options gamma. This is the behavioural analog: isolate the 223 classified dealers inside the same ≥$100k book, map inventory, and publish the liquidation-density profile plus the gamma flip.

Live BTC preview at fetch: mark $78,999.50. Dealers net short $26.7M. Gamma flip $61,965 (−21.6%). ETH dealers net short $203.6M, flip $3,146.13 (+26.5%). SOL dealers net short $52.4M, flip $82.03 (−20.9%).

When the regime flag reads amplify, fresh size is a different decision than when it reads dampen. Candles do not give you that flag.

5. Regimes — which playbook is allowed?

Short-term board: Crypto Market Regimes. Long-horizon board: Long-horizon regimes.

Endpoints: GET /api/v1/quant/market (Pro), GET /api/v1/regimes/current (any valid key).

Short-term live read at fetch: whole-market Squeeze, model confidence 98%, in regime 24h. Chance of a vol spike in 24h: 1%. Chance of a >5% drop in 24h: 4%. Direction buckets: 10% strong fall, 24% mild fall, 36% sideways, 22% mild rise, 9% strong rise.

Use the regime to pick which strategy may run. Do not use it as a magic “buy” button. The product page is blunt about that, and so am I.

6. Hyperliquid market data — funding, OI, depth, liquidations

Venue page: Hyperliquid.

Live at fetch: $10.90B open interest across 177 perp markets. BTC share of tracked-venue OI: 13.3%. BTC funding APR: −7.6%. Whale bias on that page: SHORT (45.3% long).

This is the tape after the decision layer. Funding, OI, depth, and liquidations refine size and timing. They do not replace calendar → flows → whales → GEX → regimes.

Execution skill vs data layer

Layer What it does Example
Execution skill Talks to a venue. Places or cancels orders. Bybit EU AI Trading Skill → REST
Data layer Tells the agent what is worth trading now. CryptoDataAPI Pro boards above

Bybit solved “where can my chat agent send an order.” Bitpanda and Scalable, per Trending Topics, solved a similar problem with MCP and human confirmations.

Hyperliquid builders still need the second layer first. Without it, an agent with a key is just a fast way to be wrong.

We teach agents what to read first. We do not place trades for them.

Optional research-only checklist (no keys for trading, no orders)

Paste this into a research agent. Do not attach a venue trading key. Do not place orders.

Research-only. No orders. No venue trading keys.

1) GET https://cryptodataapi.com/api/v1/event/regime?window_days=7
   — list catalysts in the next 7 days; note Event Risk score/band.
2) GET .../market-intelligence/etf/btc/flows and .../etf/eth/flows
   — report latest print and whether the recent run is inflow or outflow.
3) GET .../quant/whales  (Pro; if 403, stop and say 403)
   — net_bias, long_pct, top 3 coins by gross, directional_net_usd where present.
4) GET .../quant/gex?symbol=BTC  (Pro; if 403, stop and say 403)
   — mm net inventory, gamma_flip, amplify/dampen if present.
5) GET .../quant/market?horizon=24h  (Pro)
   — regime label, confidence, 24h direction buckets.
6) Read https://cryptodataapi.com/exchanges/hyperliquid
   — OI, BTC funding APR, whale bias.

Return a brief: what to avoid, what is crowded, what catalyst is next.
Do not recommend a trade size. Do not invent numbers for 403s.

Get a free key and connect the read path

  1. Free key at cryptodataapi.com. Confirm the email. That lifts the same key and turns Pro on for 24 hours.
  2. Agent setup and MCP: cryptodataapi.com/ai-agents.
  3. Current plans: cryptodataapi.com/pricing — Free forever for the core live suite; Pro from $29/mo billed annually ($39/mo monthly); Pro Plus from $99/mo billed annually for history and bulk.

Machine-readable map: llms.txt. Docs: api/docs.

Bybit taught agents where to trade. We teach them what to read first.

Frequently asked questions

Does Hyperliquid have an official AI trading skill like Bybit's?

No first-party one that we could find as of 8 September 2026. What exists is third-party: community skills and wrappers. Bybit's skill is an open-source package that talks to Bybit's REST API; Hyperliquid builders still need a data layer before handing an agent an execution key.

What should a Hyperliquid AI agent read before trading?

In order: the event calendar (what hits the tape next), spot ETF flows (is institutional demand real today), classified whale positioning (who is leaning), perp gamma exposure (where dealers amplify or dampen), and the regime read (which playbook is allowed). Funding, open interest, depth and liquidations refine size and timing after that decision layer.

Which of these boards are free?

The event calendar, event regime, ETF flows, the long-horizon regime and the Hyperliquid venue page work on any valid key. Whale activity, gamma exposure and the short-term HMM regime are Pro. Confirming the email on a free key turns Pro on for 24 hours.

Does CryptoDataAPI place trades?

No. It is a read-only data layer. Agents read; they do not trade through it. Execution stays with the venue and the human.