Hyperliquid API Reference for Openclaw

No package installation is required: this skill is a compact API reference used by an AI coding agent or developer during Hyperliquid integration work.

1. Add the skill to the agent's skills directory according to the host Openclaw Skills installation process.
2. Use the official documentation root and fetch raw Markdown by appending .md to a documentation URL when current details need verification.
3. Configure the correct environment endpoints:

export HYPERLIQUID_REST_URL=https://api.hyperliquid.xyz
export HYPERLIQUID_WS_URL=wss://api.hyperliquid.xyz/ws
export HYPERLIQUID_EVM_RPC_URL=https://rpc.hyperliquid.xyz/evm

For testnet, use https://api.hyperliquid-testnet.xyz, wss://api.hyperliquid-testnet.xyz/ws, and https://rpc.hyperliquid-testnet.xyz/evm with chain 998 instead of mainnet chain 999.

4. Set the JSON content type for REST requests:

curl -X POST "$HYPERLIQUID_REST_URL/info" \
  -H "Content-Type: application/json" \
  -d '{"type":"allMids"}'

5. Load live metadata with /info before constructing orders; do not hardcode asset IDs, decimals, leverage, or tick rules.
6. Keep signing keys outside source code and use an API wallet only for permitted L1 actions. Never expose private keys in prompts, logs, generated documentation, or agent context.
7. For testnet funding, use the documented faucet at https://app.hyperliquid-testnet.xyz/drip and account for its eligibility requirements.

Reference metadata

API organization

• REST endpoints: Mainnet and testnet /info for unsigned reads and /exchange for signed actions.
• WebSocket channels: Market data, candles, order books, trades, account state, fills, funding, ledger updates, TWAP state, and notifications.
• HyperEVM: Separate RPC endpoints with chain 999 on mainnet and 998 on testnet.
• Read payloads: /info requests use a type field with request-specific parameters and return the bare payload.
• Write payloads: /exchange requests contain an action, millisecond nonce, signature, and optional vault or expiration fields; responses use ok or err status.

Trading data taxonomy

• Market metadata: meta, spotMeta, perpDexs, allPerpMetas, and asset context responses.
• Market activity: allMids, l2Book, recentTrades, candleSnapshot, funding history, and predicted funding.
• Account state: Clearinghouse state, spot balances, open orders, historical orders, fills, funding, ledger updates, portfolio, fees, and rate limits.
• Identity and permissions: User role, extra agents, subaccounts, vault details, delegations, referral data, and builder-fee permissions.
• Order vocabulary: Limit, IOC, GTC, ALO, trigger, TP/SL, reduce-only, TWAP, resting, filled, canceled, rejected, and waiting states.
• Asset identifiers: Perp indexes from live meta, spot IDs beginning at 10000, HIP-3 IDs using dex offsets, and HIP-4 outcome encodings.
• Operational constraints: Price significant figures, decimal limits, size decimals, minimum notional, open-order caps, nonce windows, pagination, and REST/WebSocket quotas.

Historical time-ranged ledger and funding queries return up to 500 items per response; continue pagination with the last returned time. Many list-style reads cap responses at 2,000 items, while candles return only the most recent 5,000 snapshots.