Bitget Agent Hub

Connect AI assistants to Bitget — trade, query, and manage your crypto portfolio through natural language.

---

Bitget Agent Hub connects AI assistants and automation tools to the Bitget exchange. Two integration modes, plus a built-in Skill Hub for market analysis:

• MCP Server — for Claude Code, Cursor, Codex, and any MCP-compatible AI
• CLI (bgc) + Skill — for shell-based AI agents (Claude Code skills, OpenClaw)

Once configured, your AI can check prices, query balances, place and cancel orders, manage futures positions, set leverage, and transfer funds — all through natural language.

---

Installation

Use bitget-hub to install, upgrade, and manage all Bitget Agent Hub packages. No installation required — run directly via npx.

Quick Start

# Install everything and deploy skills to Claude Code (default)
npx bitget-hub upgrade-all --target claude

This installs all three packages globally and deploys skills to Claude Code:

Package	What it does
bitget-client	CLI tool (bgc) — shell access to all Bitget tools
bitget-skill	Trading skill — AI uses bgc as a live API bridge
bitget-skill-hub	5 market-analysis skills (macro, sentiment, technical, news, on-chain)

Deploy Skills to AI Tools

Skills can be deployed to Claude Code, Codex, and OpenClaw:

# Deploy to a specific tool
npx bitget-hub install --target codex

# Deploy to multiple tools
npx bitget-hub install --target claude,codex

# Deploy to all supported tools
npx bitget-hub install --target all

# Deploy a single skill package
npx bitget-hub install bitget-skill --target claude

Install or Upgrade Individual Packages

# Upgrade a single package
npx bitget-hub upgrade bitget-client

# Upgrade and deploy skills
npx bitget-hub upgrade bitget-skill --target claude

# Rollback to a specific version
npx bitget-hub rollback bitget-skill --to 1.0.0

Interactive Mode

npx bitget-hub

Launches an interactive menu to upgrade, rollback, or install skills — no flags needed.

Set Credentials

All integrations need a Bitget API key for private endpoints (account, trading). Public market data works without credentials.

1. Log in to bitget.com → Settings → API Management
2. Create a new API key — select Read and/or Trade permissions
3. Set environment variables:

export BITGET_API_KEY="your-api-key"
export BITGET_SECRET_KEY="your-secret-key"
export BITGET_PASSPHRASE="your-passphrase"

---

MCP Server

Gives AI assistants direct access to Bitget tools via the Model Context Protocol. No global install needed — runs via npx.

Claude Code

claude mcp add -s user \
  --env BITGET_API_KEY=your-api-key \
  --env BITGET_SECRET_KEY=your-secret-key \
  --env BITGET_PASSPHRASE=your-passphrase \
  bitget \
  -- npx -y bitget-mcp-server

Codex

Add to ~/.codex/config.toml:

[[mcp_servers]]
name = "bitget"
command = "npx"
args = ["-y", "bitget-mcp-server"]

[mcp_servers.env]
BITGET_API_KEY = "your-api-key"
BITGET_SECRET_KEY = "your-secret-key"
BITGET_PASSPHRASE = "your-passphrase"

OpenClaw

Add to your OpenClaw agent config:

{
  "mcp_servers": {
    "bitget": {
      "command": "npx",
      "args": ["-y", "bitget-mcp-server"],
      "env": {
        "BITGET_API_KEY": "your-api-key",
        "BITGET_SECRET_KEY": "your-secret-key",
        "BITGET_PASSPHRASE": "your-passphrase"
      }
    }
  }
}

→ See docs/packages/bitget-mcp.md for more clients (Claude Desktop, Cursor, VS Code Copilot, Windsurf).

---

CLI Tool (bgc)

A command-line interface for all Bitget tools. Outputs JSON — ideal for scripting and AI agent shell use.

# Market data (no credentials needed)
bgc spot spot_get_ticker --symbol BTCUSDT

# Account queries
bgc account get_account_assets

# Trading
bgc spot spot_place_order --orders '[{"symbol":"BTCUSDT","side":"buy","orderType":"limit","price":"95000","size":"0.01"}]'

---

Skill Hub

5 market-analysis skills for Claude Code, Codex, and OpenClaw. Each skill instructs the AI how to use the market-data MCP server to deliver analyst-grade outputs.

Skill	What it does
macro-analyst	Macro & cross-asset analysis — Fed policy, yield curve, BTC vs DXY/Nasdaq/Gold
market-intel	On-chain & institutional intelligence — ETF flows, whale activity, DeFi TVL
news-briefing	News aggregation & narrative synthesis — morning briefings, keyword search
sentiment-analyst	Sentiment & positioning — Fear & Greed, long/short ratios, funding rates
technical-analysis	Technical analysis — 23 indicators across 6 categories

The technical-analysis skill requires Python: pip install pandas numpy

→ See docs/skill-hub.md for full documentation.

---

Modules

Module	Tools	Loaded by default
spot	13	✅
futures	14	✅
account	8	✅
margin	7	—
copytrading	5	—
convert	3	—
earn	3	—
p2p	2	—
broker	3	—

Default: spot + futures + account = 36 tools (fits within Cursor's 40-tool limit). Load everything: --modules all

---

Security

• Credentials via environment variables only — never hardcoded or logged
• --read-only flag disables all write operations at server level
• All authenticated requests signed with HMAC-SHA256
• Client-side rate limiting prevents accidental API abuse
• Write operations (orders, transfers) require explicit confirmation before execution

---

Development

# Prerequisites: Node.js ≥ 18, pnpm ≥ 8
pnpm install
pnpm -r build
pnpm -r test

---

License

MIT