flipper-mcp-bridge

flipper-mcp-bridge

An MCP server that exposes a USB-connected Flipper Zero device as tools for controlling infrared devices. It enables listing, parsing, and transmitting IR signals from saved files and built-in universal remotes.

Category
访问服务器

README

flipper-mcp-bridge

MCP server that exposes a USB-connected Flipper Zero as a set of tools for MCP clients (Claude Code, Home Assistant via compatible bridge, etc.). v0 focuses on IR: list/parse saved .ir files, replay buttons, capture new signals.

Tested against Momentum firmware (mntm-008). Any recent Flipper fork with the same CLI (ir tx, ir rx, storage *, loader *) should work.

v0 tools

Tool Purpose
device_info Return the Flipper's device_info dict (firmware, hardware, radio)
list_ir_files List .ir files under a directory on the SD card
list_ir_buttons Parse a saved .ir file and return its buttons
send_ir_button Transmit a named button from a saved .ir file
send_ir_signal Transmit an ad-hoc parsed IR signal by MSB-first integer hex (e.g. NECext DF02 EE11)
list_universal_remotes List built-in universal IR remotes available on the firmware (ac, tv, fans, ...)
list_universal_signals List the signal names for a built-in universal remote
send_universal_signal Transmit a named signal from a built-in universal remote
learn_ir_button Put the Flipper in RX, capture the next remote press, append it to a .ir file

Setup (Windows host, WSL2)

1. Forward the Flipper USB into WSL

Install usbipd-win on Windows:

winget install usbipd

Then (from Windows PowerShell):

usbipd list                          # find the Flipper's BUSID
usbipd bind --busid <X-Y>            # one-time, admin PowerShell
usbipd attach --wsl --busid <X-Y>    # each replug / reboot

After attach, the Flipper shows up in WSL as /dev/ttyACM0.

2. Grant serial access in WSL

sudo usermod -aG dialout $USER

Then restart WSL so the group is active:

wsl.exe --shutdown

Reopen your shell. groups should now include dialout.

3. Install dependencies

From the repo root:

uv sync

4. Smoke test

uv run python scripts/smoketest_readonly.py

You should see device info, a list of files under /ext/infrared/, and parsed contents of each .ir file.

Running

MCP stdio (for Claude Code, Cursor, etc.):

uv run flipper-mcp-bridge

HTTP REST API (for Home Assistant, curl, scripts):

uv run flipper-mcp-bridge --http --port 8765

Endpoints:

Method Path Body / Query
GET /health —
GET /device —
GET /ir/files ?dir=/ext/infrared
GET /ir/buttons ?file=/ext/infrared/Remote.ir
POST /ir/send-button {"file": "...", "button": "..."}
POST /ir/send-signal {"protocol": "...", "address": "...", "command": "..."}
GET /ir/universal/list ?remote=ac (omit to list available remotes)
POST /ir/universal/send {"remote": "ac", "signal": "OFF"}
POST /ir/learn {"file": "...", "button": "...", "timeout_seconds": 30}

Home Assistant integration

Deployment note: reachability from HA

HA needs to be able to HTTP to the bridge. Two easy setups work out of the box:

  1. Run the bridge on the same host as HA (Pi/NUC/server with Flipper plugged in). HA hits http://127.0.0.1:8765. Simplest.
  2. Run the bridge on any always-on Linux host on the LAN. Start it with --host 0.0.0.0 (the CLI prints a warning — there's no auth in v1, so only do this on a trusted LAN). HA hits http://HOST:8765.

WSL2 caveat: WSL2 uses NAT — the WSL IP isn't reachable from other hosts on the LAN. Running the bridge inside WSL2 and expecting HA on a different device to reach it requires netsh interface portproxy port-forwarding on the Windows host, or running the bridge on the Windows host directly (Python + pyserial work fine on Windows).

Configuration

Drop this into configuration.yaml:

rest_command:
  flipper_humidifier_toggle:
    url: "http://FLIPPER_HOST:8765/ir/send-button"
    method: POST
    content_type: "application/json"
    payload: '{"file":"/ext/infrared/Remote.ir","button":"Humid"}'

  flipper_ac_off:
    url: "http://FLIPPER_HOST:8765/ir/universal/send"
    method: POST
    content_type: "application/json"
    payload: '{"remote":"ac","signal":"OFF"}'

Then in automations or scripts:

action:
  - service: rest_command.flipper_humidifier_toggle

For a switch-like entity, use a RESTful switch pointed at the same /ir/send-button endpoint (state is maintained by HA since the Flipper itself doesn't expose device state).

Port selection

The bridge picks a serial device in this priority order:

  1. Explicit port= argument (library use only)
  2. FLIPPER_PORT environment variable
  3. Auto-detect: the first attached device whose USB manufacturer is "Flipper Devices Inc." (or VID:PID 0483:5740)
  4. Fallback: /dev/ttyACM0

So in the common case you don't need to set anything. If you've got multiple CDC devices and want to pin a specific one:

FLIPPER_PORT=/dev/ttyACM1 uv run flipper-mcp-bridge

Or add env to the .mcp.json server entry.

Registering with Claude Code

The repo ships a .mcp.json at the root — Claude Code picks it up automatically when you start a session in this directory (you'll be prompted to trust it on first launch). If you'd rather register it explicitly:

claude mcp add flipper -- uv run --directory "$(pwd)" flipper-mcp-bridge

Known limitations

  • Capture latency: learn_ir_button takes a few seconds before the Flipper actually starts listening. Press the remote a beat after calling the tool, not immediately.
  • Transport: CLI-over-serial only. Protobuf RPC is not wired yet. Fine for IR at human pace; may be revisited for throughput-sensitive flows.
  • Blocked by foreground apps: if a non-CLI app owns the Flipper (e.g. the Nightstand Clock idle screen), IR TX is blocked. The bridge tries loader close once on conflict, but some apps can only be exited manually on the device.
  • Raw IR capture not supported: unknown-protocol signals won't round-trip via learn_ir_button yet.
  • IR only: Sub-GHz, NFC, RFID, GPIO, BadUSB — none of these are wired.

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
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
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
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
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
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
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
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