Liquidity.io Docs

Liquidity.io provides native integration with Claude Code and any MCP-compatible AI client for agentic trading. No custom framework required — use Anthropic's standard tooling.

Quick Start

1. Install Claude Code

npm install -g @anthropic-ai/claude-code

2. Configure MCP Connection

Create .mcp.json in your project:

{
  "mcpServers": {
    "liquidity": {
      "command": "npx",
      "args": ["@liquidityio/mcp"],
      "env": {
        "LIQUIDITY_API_KEY": "your-api-key",
        "LIQUIDITY_API_SECRET": "your-api-secret",
        "LIQUIDITY_API_URL": "https://api.liquidity.io/v1"
      }
    }
  }
}

3. Trade with Natural Language

> What's the current price of AAPL?
[Calling get_quote: symbol=AAPL]
AAPL is trading at $185.42 (bid: $185.40, ask: $185.44)

> Buy 100 shares at market
[Calling place_order: symbol=AAPL, side=buy, type=market, quantity=100]
Order filled: 100 AAPL @ $185.43 avg

> Show my portfolio
[Calling get_portfolio]
AAPL: 100 shares @ $185.43 ($18,543.00)
Cash: $81,457.00

MCP Tools

The @liquidityio/mcp server exposes 8 tools:

Tool	Description
`get_quote`	Real-time bid/ask/last for any symbol
`get_orderbook`	Full depth-of-book (L2 data)
`place_order`	Submit market, limit, stop orders
`cancel_order`	Cancel open orders
`get_portfolio`	Account balances and positions
`get_open_orders`	List all open orders
`get_trades`	Trade history with P&L
`get_markets`	Available instruments and trading pairs

Trading Skills

Pre-built slash commands for Claude Code:

Skill	Command	Description
Market Analysis	`/analyze AAPL`	Technical + fundamental analysis
Portfolio Rebalance	`/rebalance`	Suggest and execute rebalancing
Risk Assessment	`/risk`	Portfolio risk metrics (VaR, Sharpe, etc.)
Order Preview	`/preview buy 100 AAPL`	Simulate order before execution
Trade History	`/history`	P&L report with breakdown

Install Skills

# Symlink into Claude Code
ln -s skills/discover-liquidity \
      ~/.claude/skills/discover-liquidity

Risk Controls

All AI trading is subject to server-side risk limits:

Control	Default	Configurable
Max order size (USD)	$100,000	Yes
Max daily loss (USD)	$10,000	Yes
Max position size (USD)	$500,000	Yes
Require confirmation	Yes (orders > $10K)	Yes
Withdrawals	Blocked	N/A
Account modifications	Blocked	N/A

Kill Switch

Revoke the API key to immediately:

Cancel all open orders
Block new order submission
Preserve existing positions

Compatible AI Clients

Any MCP-compatible client works with the Liquidity.io MCP server:

Client	Status	Notes
Claude Code	Full support	Native MCP integration
Claude Desktop	Full support	Via MCP server config
Cursor	Full support	MCP server in settings
Windsurf	Full support	MCP server in settings
Custom MCP client	Full support	Any client implementing MCP spec

Institutional Patterns

Supervised Trading

{
  "riskLimits": {
    "maxOrderSizeUSD": 1000000,
    "requireConfirmation": true,
    "confirmationThresholdUSD": 50000,
    "allowedSymbols": ["BTC-USD", "ETH-USD", "AAPL", "TSLA"],
    "allowedOrderTypes": ["limit", "market"]
  }
}

Audit Trail

Every MCP tool call is logged with:

Timestamp, tool name, parameters
User identity (API key)
Order ID (if applicable)
Confirmation status (approved/rejected)

Multi-Agent Coordination

Run multiple Claude Code instances with different strategies:

# Terminal 1: Market maker
claude "Run a market making strategy on BTC-USD with 10bp spread"

# Terminal 2: Risk monitor
claude "Monitor my portfolio and alert if daily loss exceeds $5000"

# Terminal 3: Rebalancer
claude "Rebalance to 60/40 BTC/ETH every hour"

Each instance uses the same MCP server but with independent API keys and risk limits.

AI Trading — Claude Code & MCP