msfs2024-mcp

msfs2024-mcp

Enables MCP clients like Claude to read live state and control aircraft in Microsoft Flight Simulator 2024 via SimConnect, FSUIPC7, and raw memory, offering 23 tools for simvars, events, autopilot, and more.

Category
访问服务器

README

msfs2024-mcp

✈️ Also in this repo:

  • a dark-mode PyQt6 electronic flight checklist app (Cessna 172S + Piper Archer II, normal/abnormal/emergency procedures) designed to float on top of MSFS 2024 — pip install -e ".[checklist]", run msfs-checklist.
  • a Claude-powered controls setup advisor for the Honeycomb Alpha/Bravo and VelocityOne Rudder: per-aircraft binding plans with procedure coaching, plus LLM review of your exact hardware — pip install -e ".[controls]", run msfs-controls.

A Model Context Protocol server for Microsoft Flight Simulator 2024. It lets an MCP client (Claude Desktop, Claude Code, etc.) read live sim state and drive the aircraft across three capability layers:

Layer Source Covers Setup
1. SimConnect Official Microsoft API (python-SimConnect) Thousands of SimVars, Events (controls), bundled aircraft state MSFS running
2. FSUIPC7 Offset table (fsuipc module) Values SimConnect doesn't cleanly expose; stable offsets FSUIPC7 installed + running
3. Raw memory pymem / ReadProcessMemory Escape hatch for anything else Opt-in, admin rights

Reality check: SimConnect, FSUIPC, and raw memory are Windows-only and require a running MSFS on the same machine (or LAN via SimConnect.cfg). Run this server on that Windows host. Every layer degrades gracefully — if a layer isn't available, its tools return a structured { "ok": false, "error": ... } explaining why, instead of crashing the server. So the server boots and the catalog/discovery tools work even before MSFS is up.

Install (on the Windows host with MSFS)

git clone <your-repo-url> msfs2024-mcp
cd msfs2024-mcp
python -m venv .venv
.venv\Scripts\activate
pip install -e .
copy .env.example .env   # then edit if you like

Layer prerequisites:

  • SimConnect — installed automatically with the SimConnect pip package, which ships its own SimConnect.dll. Just have MSFS running and loaded into a flight.
  • FSUIPC7 — download and run FSUIPC7 (free for basic offset access). Leave it running alongside MSFS.
  • Raw memory — set MSFS_ENABLE_MEMORY=true in .env and run the server as Administrator. Off by default.

Verify it works

python scripts\smoke_test.py

With MSFS loaded into a flight you'll see a live aircraft-state snapshot and an event round-trip (nav lights toggle). Off-Windows or with the sim closed, it prints layer health and exits cleanly — which is how you know the graceful-degradation path is intact.

Platform-independent tests (catalog integrity + degradation) run anywhere:

pip install -e ".[dev]"
pytest -q

Wire into an MCP client

Claude Desktop / Claude Code — add to your MCP config (claude_desktop_config.json or .mcp.json):

{
  "mcpServers": {
    "msfs2024": {
      "command": "python",
      "args": ["-m", "msfs_mcp.server"],
      "cwd": "C:\\path\\to\\msfs2024-mcp",
      "env": { "MSFS_ENABLE_FSUIPC": "true", "MSFS_ENABLE_MEMORY": "false" }
    }
  }
}

(If you pip install -e ., you can use the msfs-mcp console script instead of python -m msfs_mcp.server.)

Tool surface (23 tools)

Connection — connection_status, connect_sim

SimConnect / SimVars — get_simvar, get_simvars, set_simvar, get_aircraft_state

SimConnect / Events & autopilot — trigger_event, autopilot_set_heading, autopilot_set_altitude, autopilot_set_vertical_speed, autopilot_toggle_master

Discovery — list_simvars, list_events (searchable by keyword/category)

FSUIPC — fsuipc_status, fsuipc_read_offset, fsuipc_read_known, fsuipc_write_offset

Raw memory — memory_status, memory_attach, memory_module_base, memory_read, memory_read_pointer_chain, memory_write

Resources — msfs://telemetry/state, msfs://catalog/simvars, msfs://catalog/events

Prompts — preflight_briefing, fly_to_heading_altitude

Examples (natural language to the MCP client)

  • "What's my current altitude and heading?" → get_aircraft_state
  • "Find me every autopilot-related variable." → list_simvars(category="autopilot")
  • "Raise the landing gear and set flaps to the first notch." → trigger_event('GEAR_UP'), trigger_event('FLAPS_INCR')
  • "Engage the autopilot for heading 270 at 8000 feet." → autopilot_toggle_master, autopilot_set_heading(270), autopilot_set_altitude(8000)
  • "Read FSUIPC offset 0x0560 as a long." → fsuipc_read_offset('0x0560', 'l')

Safety notes

  • Writes are real. set_simvar, trigger_event, and fsuipc_write_offset change the running sim. Fine for your own local flight; think before scripting them.
  • Raw memory is double-gated. It requires both MSFS_ENABLE_MEMORY=true and allow_write=true per write call, because bad writes can crash MSFS. Pointer chains break on most sim updates — keep them version-pinned.
  • Reads are harmless. All read paths are observation-only.

Extending the catalog

src/msfs_mcp/catalog.py is a curated subset, not the full SDK. The generic get_simvar / set_simvar / trigger_event tools reach any SimVar or Event by exact SDK name — add entries to the catalog only to make them discoverable. Full reference: the MSFS SDK SimVars and Event IDs docs.

Architecture

MCP client  ──stdio──▶  msfs_mcp.server (FastMCP, 23 tools)
                              │
              ┌───────────────┼────────────────┐
              ▼               ▼                 ▼
     SimConnectClient   FsuipcClient      MemoryClient
     (SimConnect.dll)   (FSUIPC7)         (pymem)
              │               │                 │
              └──────── Microsoft Flight Simulator 2024 ────────┘

Each client is a singleton with lazy connect, a thread lock around the native handle, and a uniform LayerUnavailable error contract that the server renders as structured JSON.

License

MIT

Companion apps

  • EFB Checklist — PyQt6 always-on-top electronic checklist (normal + emergency procedures, V-speeds) for flying alongside MSFS 2024. Ships with the Cessna 172S and Piper PA-28-181 Archer.

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
Kagi MCP Server

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。

官方
精选
Python
graphlit-mcp-server

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。

官方
精选
TypeScript
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

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。

官方
精选
mcp-server-qdrant

mcp-server-qdrant

这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。

官方
精选