ibkr-mcp
Read-only MCP server for Interactive Brokers that exposes market data, positions, and account info as MCP tools.
README
ibkr-mcp
Read-only MCP server for Interactive Brokers Gateway via the TWS socket API. Connects directly to your running IB Gateway on localhost — no Client Portal REST API, no bundled Java gateway, no 264 MB npm packages.
What it does
Exposes IB Gateway market data, positions, and account info as MCP tools that any MCP client (Claude Code, Claude Desktop, etc.) can call.
Read-only by design. No order placement tools. The connection uses readonly=True at the API level — IB Gateway will reject order submissions even if the code is modified.
Prerequisites
- IB Gateway or Trader Workstation (TWS) running on localhost (default port 4001)
- Python 3.11+
- An active IBKR account (paper or live)
Installation
git clone https://github.com/mark-liu/ibkr-mcp.git
cd ibkr-mcp
python3 -m venv .venv
source .venv/bin/activate
pip install -e .
Configuration
All configuration is via environment variables:
| Variable | Default | Description |
|---|---|---|
IB_HOST |
127.0.0.1 |
Gateway host |
IB_PORT |
4001 |
Gateway port (4001=live, 4002=paper) |
IB_CLIENT_ID |
10 |
API client ID (must be unique per connection) |
IB_MARKET_DATA_TYPE |
3 |
1=live, 2=frozen, 3=delayed, 4=frozen-delayed |
IB_RECONNECT_INTERVAL |
30 |
Seconds between reconnect attempts |
IB_CACHE_TTL |
3600 |
Contract cache TTL in seconds |
Claude Code Integration
Add to ~/.claude.json:
{
"mcpServers": {
"ibkr": {
"command": "/path/to/ibkr-mcp/.venv/bin/python",
"args": ["-m", "ibkr_mcp"],
"env": {
"IB_PORT": "4001",
"IB_CLIENT_ID": "10"
}
}
}
}
Then in Claude Code, tools like ibkr_quote, ibkr_positions, ibkr_historical_bars become available automatically.
Available Tools
| Tool | Description | Key Parameters |
|---|---|---|
ibkr_quote |
Current price quotes | symbols (comma/space separated, max 20) |
ibkr_historical_bars |
OHLCV historical bars | symbol, duration ("1 M"), bar_size ("1 day") |
ibkr_positions |
Portfolio positions with P&L | — |
ibkr_account_summary |
NLV, cash, margin, buying power | — |
ibkr_option_chain |
Available expirations and strikes | symbol, exchange (optional) |
ibkr_contract_search |
Fuzzy search for contracts | pattern |
ibkr_fx_rate |
Live FX rate | pair ("EURUSD", "AUD/USD") |
ibkr_connection_status |
Gateway health check | — |
MCP Resources
| URI | Description |
|---|---|
portfolio://positions |
Current positions as context |
account://summary |
Account summary as context |
Design Decisions
- TWS socket API, not Client Portal REST. Direct connection to IB Gateway on port 4001 via
ib_async. Sub-millisecond local latency, streaming-capable, full options support. No HTTP indirection through a Java gateway. - Persistent connection with background reconnect. If IB Gateway restarts, the server automatically reconnects without manual intervention.
- Contract caching. Qualified contracts (with populated
conId) are cached for 1 hour, eliminating redundant API round-trips. - Market hours detection. Uses
exchange_calendars(NYSE) to automatically switch between live (type 1) and delayed (type 3) market data. - NaN handling. IB returns
float('nan')for missing data. Every numeric field is cleaned toNonebefore JSON serialization. - Rate limiting. Token bucket limiters respect IB's API limits: 45 req/s for market data, 1 req/s for historical data.
- Graceful degradation. Response cache stores last-known-good data, so tools return stale results (flagged) instead of errors during brief disconnects.
Running Tests
pip install -e ".[dev]"
pytest tests/ -v
Tests run without a live IB Gateway — all IB interactions are mocked.
Project Structure
src/ibkr_mcp/
__init__.py
__main__.py # Entry point (nest_asyncio + mcp.run)
server.py # FastMCP server, lifespan, tool registration, resources
client.py # IBKRClient: connection, caching, all data methods
config.py # Environment variable configuration
cache.py # Contract cache + response cache
models.py # Pydantic input validation
utils.py # NaN handling, rate limiter, retry, formatting
tools/
market.py # ibkr_quote, ibkr_historical_bars, ibkr_fx_rate
account.py # ibkr_positions, ibkr_account_summary
options.py # ibkr_option_chain
search.py # ibkr_contract_search
status.py # ibkr_connection_status
Acknowledgments
This project was built after evaluating six existing IBKR MCP servers. While none were suitable as-is (wrong API, security issues, proprietary licenses, abandoned), each contributed patterns and lessons:
- xiao81/IBKR-MCP-Server (Apache-2.0) — FastMCP lifespan pattern with typed context, MCP resources for portfolio/account data
- ArjunDivecha/ibkr-mcp-server (MIT) — Rate limiting and retry decorator patterns, symbol validation approach, exception hierarchy design
- omdv/ibkr-mcp-server — Market hours detection via
exchange_calendars, contract caching concept, market data type switching - jeffbai996/ibkr-terminal — Background reconnect loop concept, cached degradation pattern, NaN handling throughout, subscription cleanup patterns
- code-rabi/interactive-brokers-mcp (MIT) — Tool definition and registration patterns, read-only mode enforcement approach
- rcontesti/IB_MCP (MIT) — Endpoint categorization and tool description patterns
No code was copied from any of these projects. All implementations are original.
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。