Use with Cursor and Claude Code

HookKit ships a hosted Model Context Protocol server at https://api.hookkit.dev/mcp. Any MCP client (Cursor, Claude Code, Claude Desktop, Windsurf, and others) can use HookKit as native tools with a one-line config. Nothing to install, no API key.

Cursor

One-click install: Add HookKit to Cursor

Or add to .cursor/mcp.json in your project (or ~/.cursor/mcp.json globally):

{
  "mcpServers": {
    "hookkit": {
      "url": "https://api.hookkit.dev/mcp"
    }
  }
}

Claude Code

claude mcp add --transport http hookkit https://api.hookkit.dev/mcp

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "hookkit": {
      "serverUrl": "https://api.hookkit.dev/mcp"
    }
  }
}

Available tools

Tool	What it does
`decode_hook`	Permissions encoded in a hook address, with capability notes. Pure, no chain call
`inspect_hook`	Full profile of a deployed hook: bytecode, proxy, warnings, pools
`list_hooks`	Every hook used by a v4 pool, by pool count or recency
`mine_hook_salt`	CREATE2 salt for a hook address with the requested permissions
`list_pools`	Indexed v4 pools, filterable by hook, token, hooked/unhooked
`get_pool`	Live pool state: price, tick, liquidity, LP fee, hook permissions
`compute_pool_id`	PoolId for a pool key. Pure, no chain call
`quote`	Quote through the pool’s hook via the V4Quoter
`build_swap`	Unsigned approval + UniversalRouter transactions to sign yourself
`get_stats`	v4 contract addresses and index totals

See the Agent API for prompts and protocol details.

Non-custodial by design

build_swap returns unsigned calldata only. Your agent or editor session signs with its own wallet; HookKit never sees keys and never submits transactions.