FIX MCP Server
A production-grade MCP server for FIX protocol trading operations that enables order management, session repair, and algorithmic execution. It provides specialized tools for monitoring session health, managing ticker reference data, and executing complex trading scenarios.
README
FIX MCP Server
A production-grade MCP server for FIX protocol trading operations. Exposes 15 tools, 4 resources, and 6 role-specific prompts over stdio. Backed by a full PostgreSQL + Redis stack for production deployment.
What it includes
- 15 MCP tools — order management, FIX session repair, ticker operations, algo execution, scenario loading
- 13 trading scenarios — 24-hour coverage from 02:05 ET pre-dawn through 16:32 ET after-hours
- 6 role prompts — specialized system prompts for trading-ops, session-engineer, order-desk, ticker-ops, risk-compliance, algo-trader
- Algorithmic order engine — TWAP, VWAP, POV, IS, DARK_AGG, ICEBERG with schedule tracking and execution quality metrics
- Production stack — PostgreSQL 16 + Redis 7 via Docker Compose, async FIX TCP connector
Quick Start (local)
# Install into local venv
XDG_CACHE_HOME=/tmp ./.venv/bin/python -m pip install -e . --no-build-isolation
# Start the MCP server (stdio)
SCENARIO=morning_triage ./.venv/bin/fix-mcp-server
Quick Start (Docker)
docker compose up -d
# MCP stdio: docker compose run --rm mcp-server
# REST API: http://localhost:8000
# Dashboard: http://localhost:8787
MCP Client Configuration
{
"mcpServers": {
"fix-mcp": {
"command": "/home/urbano-claw/FIX-MCP/fix-mcp-server/.venv/bin/fix-mcp-server",
"args": [],
"cwd": "/home/urbano-claw/FIX-MCP/fix-mcp-server"
}
}
}
Environment Variables
| Variable | Default | Description |
|---|---|---|
SCENARIO |
morning_triage |
Scenario to load on startup |
FIX_MCP_CONFIG_DIR |
auto-discovered | Config directory override |
DATABASE_URL |
— | PostgreSQL connection string (production) |
REDIS_URL |
— | Redis connection string (production) |
LOG_LEVEL |
INFO |
Logging level |
Tools (15)
| Tool | Category | Description |
|---|---|---|
query_orders |
Orders | Query OMS with filters; returns notional and SLA countdowns |
check_fix_sessions |
Sessions | Session health, seq numbers, heartbeat age, latency |
send_order |
Orders | NewOrderSingle via FIX 35=D with auto-routing |
cancel_replace |
Orders | Cancel (35=F) or replace (35=G) an existing order |
check_ticker |
Reference | Symbol/CUSIP lookup with corporate actions and open order count |
update_ticker |
Reference | Rename symbol and bulk-update all open orders |
load_ticker |
Reference | Load new symbol and release pending IPO orders |
fix_session_issue |
Sessions | Resend request, sequence reset, or reconnect |
validate_orders |
Orders | Pre-flight validation against symbol, venue, and compliance rules |
run_premarket_check |
Ops | Full pre-market health check across all systems |
send_algo_order |
Algos | Submit TWAP/VWAP/POV/IS/DARK_AGG/ICEBERG parent order |
check_algo_status |
Algos | Schedule deviation, IS shortfall, execution quality |
modify_algo |
Algos | Pause, resume, or update POV participation rate |
cancel_algo |
Algos | Cancel algo and send cancel for all child slices |
list_scenarios |
Scenarios | List or load trading scenarios |
See docs/tools.md for full parameter schemas.
Scenarios (13)
| Name | Time | Key Problems |
|---|---|---|
morning_triage |
06:15 ET | ARCA down, BATS seq gap, ticker rename ACME→ACMX |
bats_startup_0200 |
02:05 ET | BATS SequenceReset with unexpected NewSeqNo |
predawn_adrs_0430 |
04:35 ET | Shell ADR rebrand RDSA→SHEL, ARCA latency 220ms |
preopen_auction_0900 |
09:02 ET | MOO imbalance, IEX feed stale |
open_volatility_0930 |
09:35 ET | GME LULD halt, BATS packet loss |
venue_degradation_1030 |
10:32 ET | NYSE latency 180ms, Mahwah route flap |
ssr_and_split_1130 |
11:34 ET | RIDE SSR trigger, AAPL 4:1 split in 26 min |
iex_recovery_1400 |
14:03 ET | IEX recovered, D-Limit rerouting |
eod_moc_1530 |
15:31 ET | MOC cutoff, GTC preservation |
afterhours_dark_1630 |
16:32 ET | Dark pool offline, Liquidnet SessionStatus=8 |
twap_slippage_1000 |
10:05 ET | TWAP behind schedule, GME halted mid-algo |
vwap_vol_spike_1130 |
11:35 ET | VWAP over-participation, BATS latency spike |
is_dark_failure_1415 |
14:15 ET | IS high shortfall, dark aggregator no fills |
See docs/scenarios.md for full scenario details.
Role Prompts (6)
| Prompt Name | Scope |
|---|---|
trading-ops |
General — all domains, primary triage |
session-engineer |
FIX transport layer only |
order-desk |
Routing, execution, SLA management |
ticker-ops |
Reference data, corporate actions, splits |
risk-compliance |
SSR, LULD, large order review, EOD |
algo-trader |
TWAP, VWAP, POV, IS, dark aggregator |
See docs/prompts.md for full prompt descriptions and tool assignments.
Tests
PYTHONPATH=src ./.venv/bin/python -m pytest
16 tests covering tools, scenarios, algo orders, and session repair.
CLI Smoke Test
cd /home/urbano-claw/FIX-MCP/fix-mcp-server && PYTHONPATH=src ./.venv/bin/python -c "import asyncio; from fix_mcp import server; print(asyncio.run(server.call_tool('run_premarket_check', {}))[0].text)"
Dashboard
PYTHONPATH=src ./.venv/bin/fix-mcp-dashboard
# Open: http://127.0.0.1:8787
The dashboard is self-contained — serves both the HTML and the API from one process. Features:
- Per-scenario guided workflow steps (all 13 scenarios have specific steps)
- Session health cards with latency and sequence gap indicators
- Tabbed view: Output / Sessions / Orders / Algos
- Inline algo Pause and Cancel buttons
- Send Order and Session Repair forms in sidebar
- Scenario dropdown reloads runtime and refreshes workflow steps
REST API
PYTHONPATH=src ./.venv/bin/fix-mcp-api
# Endpoints: http://localhost:8000
Standalone REST API for external integrations (OMS hooks, Claude.ai, monitoring):
GET /health— liveness probeGET /api/status— full status: scenario, sessions, orders, algosGET /api/scenarios— list all 13 scenarios with context stringsPOST /api/tool— call any MCP tool:{"tool":"...", "arguments":{...}}POST /api/reset— load a scenario:{"scenario":"morning_triage"}
Documentation
- Architecture — component diagram, data flow, engine layer
- Tools — all 15 tools with parameter schemas and examples
- Scenarios — all 13 scenarios with time, problems, and flags
- Prompts — all 6 role prompts with scope and escalation paths
- Algo Suite — algo types, execution flags, quality thresholds
- Production — Docker deploy, DB schema, FIX gateway, go-live checklist, next building blocks
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。