Kraken CLI

The official AI-native CLI for trading crypto, stocks, forex, and derivatives on Kraken. Built for agents — structured JSON output, consistent error envelopes, built-in MCP server, and paper trading with no API key required.

Repo: https://github.com/krakenfx/kraken-cli Docs: See AGENTS.md, CONTEXT.md, and skills/ in the repo for agent integration guides.

Install

See the official install instructions in the repo. Single binary, no runtime dependencies.

Verify install:

kraken --version

What It Covers

Paper Trading (No API Key Needed)

Safe sandbox against live prices — no account, no real money:

kraken paper init --balance 10000 -o json
kraken paper buy BTCUSD 0.01 -o json
kraken paper status -o json
kraken paper history -o json
kraken paper reset -o json

Futures paper trading also available: kraken futures paper

Live Market Data (No API Key Needed)

kraken ticker BTCUSD -o json
kraken orderbook BTCUSD --count 10 -o json
kraken ohlc BTCUSD --interval 60 -o json

MCP Server

Built-in Model Context Protocol server over stdio — works with Claude, Codex, Gemini CLI, Cursor, and OpenClaw:

kraken mcp                    # market + account + paper (default)
kraken mcp -s market,paper    # read-only, no auth needed
kraken mcp -s all             # all services (dangerous calls require acknowledged=true)

Add to your MCP client config:

{
  "mcpServers": {
    "kraken": {
      "command": "kraken",
      "args": ["mcp", "-s", "market,paper"]
    }
  }
}

Authentication (Live Trading)

# Interactive setup
kraken setup

# Or via environment variables
export KRAKEN_API_KEY="your-key"
export KRAKEN_API_SECRET="your-secret"

Get API keys at: https://www.kraken.com (Settings > API)

⚠️ Never pass keys directly on the CLI — use kraken setup or environment variables to avoid shell history exposure.

Agent Integration Tips

• Always use -o json — stdout is always valid JSON on success
• Route on the error field in error envelopes, not message
• Paper trading is the right starting point for any new agent setup
• The repo's skills/ folder contains 50 goal-oriented workflow packages
• AGENTS.md and CONTEXT.md in the repo cover rate limits, error handling, and invocation patterns