agent_hub
Official Bitget Agent Hub — open-source AI toolkit connecting Claude, Cursor, Codex & ChatGPT to Bitget crypto trading (89 UTA v3 ops) and market analysis. Central installer for CLI, MCP, SDK, skills & signals.
Documentation
Bitget Agent Hub — Official Bitget AI Trading and Market Data Ecosystem
The official open-source toolkit connecting Claude, Cursor, Codex, ChatGPT, and any MCP-compatible AI agent to your Bitget account and market data.
·
·
·
·
·
·
·
Overview
Bitget Agent Hub is Bitget's official AI Agent toolkit, covering cryptocurrency and tokenized US stock trading — bringing 89 Bitget UTA v3 operations across trading and market analysis to Claude, Cursor, Codex, ChatGPT, and other AI agents.
Built for individual traders, quant developers, and anyone who wants to operate a crypto exchange through natural language.
> This repository is the central entry point for the entire ecosystem. See individual package repos for detailed documentation.
Quick Start
What Do You Want to Do?
What do you want to do with AI?
|
+-- Let an AI Agent trade on Bitget for me
| |
| +-- Terminal AI (Claude Code / Codex CLI / OpenClaw)
| | -> install agent-cli + agent-skill
| |
| +-- Desktop AI (Claude Desktop / Cursor / Continue)
| -> install agent-mcp
|
+-- Let an AI Agent monitor and analyze the market for me
| -> install bitget-signal (no account, no API key)
|
+-- I'm a developer building my own Bitget AI tool
-> install bitget-agent-sdkInstall with Your AI
The use-case tree above tells you which package you need. To set up the terminal tools in one shot, paste this into your terminal AI (Claude Code, Codex, OpenClaw):
Please run `npx @bitget-ai/bitget-agent-installer upgrade-all --target all` to set up the Bitget Agent Hub terminal tools (requires Node.js 20+), then help me verify the install.This installs the `bgc` CLI plus the trading and market-signal skills, and deploys the skills to Claude Code, Codex, and OpenClaw. Desktop AI (Cursor, Claude Desktop, ChatGPT) uses `bitget-agent-mcp` instead — see Trading Tools. For per-package commands, rollback, and flags, see Installer below.
Ecosystem
Bitget Agent Hub consists of 5 core packages, each with a single responsibility. This repo also ships the `bitget-agent-installer` meta-tool that installs and upgrades them together.
Trading Tools (API Key Required)
| Package | Best For | What It Does |
|---|---|---|
| `bitget-agent-sdk` | Developers | Foundation TypeScript SDK with 89 UTA v3 operations |
| `bitget-agent-cli` | Terminal AI | `bgc` command-line tool for Claude Code, Codex, OpenClaw |
| `bitget-agent-mcp` | Desktop AI | MCP server for Claude Desktop, Cursor, Windsurf, ChatGPT |
| `bitget-agent-skill` | Terminal AI | AI reasoning guide teaching how to use `bgc` correctly |
Market Tools (No Account Needed)
| Package | What It Does |
|---|---|
| `bitget-signal` | 5 market analysis skills — macro, on-chain, sentiment, technical, news |
All packages are MIT-licensed and free to use. They can be installed side by side without conflicts.
Installer
`@bitget-ai/bitget-agent-installer` is the ecosystem's meta-tool. One command installs, upgrades, rolls back, and deploys the Bitget Agent Hub packages into your AI tools — no installing each package and running each skill's setup by hand.
npx @bitget-ai/bitget-agent-installerRun it with no arguments for an interactive menu, or pass a command directly (see below).
What It Manages
| Package | Role |
|---|---|
| `bitget-agent-cli` | The `bgc` terminal trading CLI |
| `bitget-agent-skill` | Trading reasoning guide for terminal AI |
| `bitget-signal` | Market-analysis skills (no API key) |
> Desktop AI uses `bitget-agent-mcp`, which is registered in your MCP client's config rather than installed globally — set it up via the MCP package directly. `bitget-agent-sdk` is a library dependency and needs no separate install.
Why Use It
- One command for the whole ecosystem — install, upgrade, or roll back every package together.
- Always up to date — resolves each package's latest published version automatically; nothing to pin, so differing versions across packages just work.
- Skill deployment built in — wires skills straight into Claude Code, Codex, and OpenClaw via `--target`.
- Reversible — roll any package back to a previously published version.
- Safe to preview — `--dry-run` prints every command without running it.
- Lightweight — a single dependency-free script run with `npx`; auto-detects npm or pnpm.
Commands
| Command | What It Does |
|---|---|
| _(no arguments)_ | Interactive menu — upgrade, rollback, or deploy skills |
| `upgrade-all [--target ]` | Install or update all managed packages to latest |
| `upgrade [--target ]` | Install or update one package to latest |
| `rollback --to ` | Switch a package to a specific published version |
| `install [pkg] [--target ]` | Deploy an already-installed package's skills to your AI tools |
Flags
- `--target ` — where to deploy skills: `claude`, `codex`, `openclaw`, `all` (comma-separated; default `claude`)
- `--dry-run` — preview commands without executing them
- `--version` · `--help`
Examples
# Install everything and deploy skills to every supported AI tool
npx @bitget-ai/bitget-agent-installer upgrade-all --target all
# Update just the CLI to the latest version
npx @bitget-ai/bitget-agent-installer upgrade @bitget-ai/bitget-agent-cli
# Roll the CLI back to the pre-v3 release
npx @bitget-ai/bitget-agent-installer rollback @bitget-ai/bitget-agent-cli --to 1.2.0
# Deploy already-installed market-signal skills into Claude Code and Codex
npx @bitget-ai/bitget-agent-installer install @bitget-ai/bitget-signal --target claude,codex
# Preview an upgrade without changing anything
npx @bitget-ai/bitget-agent-installer upgrade-all --target all --dry-runTrading Tools
Let an AI Agent Trade Bitget for You Automatically
Bitget Agent Hub's trading tools expose 89 UTA v3 operations through 14 curated intent verbs:
| Ask Your AI | What It Does |
|---|---|
| "Buy 0.1 BTC at market price" | Spot market order |
| "Open a BTC long, 10x leverage" | Futures position |
| "Check my balance" | Account overview |
| "Transfer USDT to futures" | Internal fund transfer |
| "Cancel all open orders" | Batch cancel |
Choose Your AI Environment
| Your AI Lives In… | Package Installed |
|---|---|
| Terminal AI: Claude Code, Codex CLI, OpenClaw | `bitget-agent-cli` + `bitget-agent-skill` |
| Desktop AI: Claude Desktop, Cursor, Continue, Windsurf, ChatGPT Desktop | `bitget-agent-mcp` |
Get Started
1. Create a Bitget API Key at bitget.com/api-management
2. Run the installer wizard or follow individual package docs
3. Verify installation by asking your AI: *"What Bitget tools are available?"*
> First time? Create a Demo API Key and use `--paper-trading` to practice before touching real funds.
Market Tools
Let an AI Agent Monitor and Analyze the Market in Real Time
`bitget-signal` adds 5 market analysis skills — no Bitget account, no API key required:
| Ask Your AI | Skill |
|---|---|
| "How is Fed policy affecting BTC?" | `macro-analyst` |
| "Are whales moving assets on-chain?" | `market-intel` |
| "Is the market fearful or greedy?" | `sentiment-analyst` |
| "Is BTC RSI overbought?" | `technical-analysis` |
| "What's important in crypto today?" | `news-briefing` |
Get Started
npx @bitget-ai/bitget-signal --target allRestart your AI tool, then ask about market conditions.
Coming Next
Bitget-exclusive signals:
- `top-trader-flow`: Top copy-trading leader positioning
- `derivatives-structure`: Perp basis and term structure
- `large-flow-detect`: Whale-sized order detection
Architecture
graph TD
subgraph Surfaces
A[bitget-agent-cliTerminal AI]
B[bitget-agent-mcpDesktop AI MCP]
C[bitget-agent-skillAI Reasoning Guide]
end
subgraph Foundation
D[bitget-agent-sdkTypeScript SDK]
end
subgraph MarketAnalysis
E[bitget-signal5 Analysis Skills]
end
A --> D
B --> D
C --> A
D --> F[Bitget UTA v3 API]
E --> G[Public Market-Data MCPno API key]
style D fill:#f9f,stroke:#333,stroke-width:2px
style F fill:#bbf,stroke:#333,stroke-width:2pxHow it works:
1. SDK (`bitget-agent-sdk`): Foundation layer with 89 operations, HMAC signing, rate limiting
2. CLI (`bitget-agent-cli`): Terminal command-line tool built on SDK
3. MCP (`bitget-agent-mcp`): Desktop AI server built on SDK
4. Skill (`bitget-agent-skill`): Teaches terminal AI how to use CLI correctly
5. Signal (`bitget-signal`): Independent market analysis via a public, no-key market-data MCP service
Supported Operations
Based on Bitget UTA v3 API — 89 operations via 14 curated intent verbs:
| Module | Default | Intent Verbs | Requires API Key |
|---|---|---|---|
| market | ✅ | `market` | No (public data) |
| trade | ✅ | `order`, `position`, `strategy_order` | Yes |
| account | ✅ | `account_overview`, `transfer_funds`, … | Yes |
| strategy | — | via `raw` | Yes |
| cryptoloans | — | `loan` | Yes |
| tax | — | `tax` | Yes |
`strategy_order` is gated on the trade module — placing strategy orders is a trading capability an agent expects whenever trading is enabled; the `strategy` module's other operations remain reachable through `raw`.
Plus meta tools: `discover` (explore operations) and `raw` (call any operation by `operationId`).
Browse the live catalog anytime with `bgc discover` (or `discover` in any surface).
Security
Credential Protection
- ✅ Never leaves your machine: API keys read from environment variables only
- ✅ Local signing: All authenticated requests signed with HMAC-SHA256 in-process
- ✅ No `.env` parsing: Tools do not read `.env` files to prevent leakage
- ✅ No proxy: Direct connection to Bitget's official API
Safety Modes
- `--read-only`: Removes all write tools at startup — AI cannot place orders or transfers
- `--paper-trading`: Routes to Bitget's Demo environment — no real funds involved
- Rate limiting: Built-in client-side throttling prevents API abuse
Best Practices
1. Use Demo API Key first for testing
2. Start with `--read-only` mode to verify queries
3. Never commit credentials to version control
4. Rotate API keys every 90 days
Documentation
| Topic | Doc |
|---|---|
| Getting started | docs/getting-started.md |
| Ecosystem architecture | docs/architecture.md |
Per-package references — CLI flags, MCP client config, SDK API, signal skills — live in each package's own repo (see the Ecosystem table above). The live tool surface is always browsable with `bgc discover` (or `discover` in any surface).
Links
- Bitget: bitget.com
- API documentation: bitget.com/api-doc
- Model Context Protocol: modelcontextprotocol.io
- All Bitget AI repos: github.com/Bitget-AI
FAQ
What is Bitget Agent Hub?
Bitget Agent Hub is the official open-source AI Agent ecosystem from Bitget. It connects Bitget's cryptocurrency trading capabilities and real-time market data to AI agents like Claude, Cursor, Codex, and ChatGPT — covering 89 UTA v3 operations through 14 curated intent verbs, plus no-account-required market analysis skills.
How do I know which package to install?
It depends on your AI tool:
- Terminal AI (Claude Code, Codex) → `bitget-agent-cli` + `bitget-agent-skill`
- Desktop AI (Claude Desktop, Cursor) → `bitget-agent-mcp`
- Market analysis only → `bitget-signal` (no API key needed)
- Developer building custom tools → `bitget-agent-sdk`
Can I use AI to trade crypto automatically?
Yes. Once installed, you can tell supported AI tools to "buy 0.1 BTC" or "open a BTC long at 10x leverage" and the AI will execute directly on your Bitget account. Use `--paper-trading` mode to rehearse safely.
Is it safe to connect my API key?
Yes. Credentials are:
- Read from environment variables only
- Never proxied through any server
- Signed locally with HMAC-SHA256
- Never logged or written to disk
How do I prevent accidental orders?
Add `--read-only` at startup. All write tools are removed entirely — the AI cannot see or call them.
Which AI tools are supported?
Claude Desktop, Claude Code, Cursor, Codex CLI, Continue, Windsurf, ChatGPT Desktop, OpenClaw, and any MCP-compatible client.
Is it free?
All packages are MIT-licensed and free to use. `bitget-signal` is maintained by Bitget and free to use — no API key, no account.
How do I update everything?
Run `npx @bitget-ai/bitget-agent-installer` and choose upgrade, or:
npx @bitget-ai/bitget-agent-installer upgrade-all --target allContributing
Issues and pull requests are welcome.
- Report bugs or request features: GitHub Issues
- Security issues: email security@bitget.com (please do not post publicly)
License
MIT License — Free for personal and commercial use. Attribution appreciated but not required.
⚠️ Risk Disclaimer: Trading cryptocurrency and tokenized stocks carries substantial risk. You are solely responsible for any orders your AI agent places on your behalf. Always use `--read-only` and `--paper-trading` modes to rehearse safely before going live.
Bitget Agent Hub is an official Bitget open-source project · Part of the Bitget AI ecosystem · Built for traders, developers, and AI enthusiasts
Frequently asked questions
What is agent_hub?
agent_hub is Official Bitget Agent Hub — open-source AI toolkit connecting Claude, Cursor, Codex & ChatGPT to Bitget crypto trading (89 UTA v3 ops) and market analysis. Central installer for CLI, MCP, SDK, skills & signals.
How do I install agent_hub?
Open the GitHub repository and follow its README. Most MCP servers are added to your client's MCP config, then called by your agent.
Is agent_hub open source?
Yes — it is hosted on GitHub at https://github.com/Bitget-AI/agent_hub and has 223 stars.
Related MCP tools
Official remote MCP server for Atlassian. Securely connect Jira, Confluence, Jira Service Management, Bitbucket, and Compass to Claude, ChatGPT, Cursor, VS Code, and other AI tools using OAuth 2.1 or API tokens.
A powerful Zotero AI and MCP plugin with ChatGPT, Gemini 3.7, Claude Fable 5, Claude Opus 5, DeepSeek V4, Grok, OpenRouter, Kimi k3, GLM 5.3, SiliconFlow, GPT-oss, Gemma 4, Qwen 3.8
The go-to web for your AI coding agent — local-first search, fetch, crawl & research over MCP. No API keys, no cloud, $0/query. Public beta.
🔥 Official Firecrawl MCP Server - Adds powerful web scraping and search to Cursor, Claude and any other LLM clients.
📦 Repomix is a powerful tool that packs your entire repository into a single, AI-friendly file. Perfect for when you need to feed your codebase to Large Lan...
MCP server to provide Figma layout information to AI coding agents like Cursor
Run your own MCP server? See who uses it and what to fix.
Measure it with TrackMCP