MCP usage

The SDK ships an stdio MCP server built on the official @modelcontextprotocol/sdk. Any agent can read public Hyperliquid data, simulate against a supplied book and check receipts. It is keyless and has no order tool.

The 12 tools

Registered tool names, in list order:

Tool	What it does
`markets`	List perpetual markets from the public info endpoint.
`order_book`	Read a capped order book and its data age.
`funding`	Read funding, mark and open interest for one coin.
`positions`	Read positions for one address.
`fees`	Read the address fee rates and their data age.
`simulate`	Simulate a closed-bar decision on a supplied book. Result is `SIMULATED`.
`funding_extremes`	Z-score of the public hourly funding rate against a trailing window.
`book_imbalance`	Bid depth divided by ask depth inside a public-book band.
`get_limits`	Read the owner limits and the remaining headroom.
`trigger_kill_switch`	Latch the kill switch. A tool call cannot clear it.
`get_cost_breakdown`	Recompute the execution-cost breakdown from a simulation receipt. Result is `SIMULATED`.
`verify_receipt`	Check a simulation receipt against its manifest and inputs bytes: verified and reproducible.

submit_order and any other order tool is refused.

Connect a client

Run cd ts/mcp-server && npm ci once in your clone (Node 22.6 or newer). Then add this block, with the absolute path to your clone:

Claude Desktop

{
  "mcpServers": {
    "perpbrain": {
      "command": "node",
      "args": ["/absolute/path/to/perpbrain-sdk/bin/perpbrain-mcp.mjs"],
      "env": {
        "PERPBRAIN_HL_NETWORK": "mainnet"
      }
    }
  }
}

Cursor

In Cursor Settings → MCP, add a server with the same shape (stdio, no secrets).

A first prompt flow

Call get_limits and note the headroom.
Call simulate on a supplied book. The result is labelled SIMULATED.
Call get_cost_breakdown on the receipt for fees, slippage and funding.
Call verify_receipt. A match reports VERIFIED.

The offline version of this flow is node --experimental-strip-types examples/mcp-fixtures.ts.

Agent-first, owner-controlledThe agent reads, simulates and checks. Only you set limits, and only you clear a kill switch. Limits you set; losses are still possible.