room-eq-wizard-mcp
Enables natural language control of Room EQ Wizard for acoustic measurements, analysis, and EQ optimization, with peak/null detection and room-mode analysis.
README
room-eq-wizard-mcp
An MCP server that connects LLMs to Room EQ Wizard (REW): typed control of REW's HTTP API plus local acoustic analysis that reduces measurement data to something a model can reason about.
REW is the DSP engine — sweeps, RT60, averaging, EQ optimization all run inside REW, and this server drives them through the API. What the server adds is the layer REW doesn't have: peak/null detection with Q factors and severity, per-band statistics, before/after comparison, room-mode prediction correlated against measurements, and log-decimated curves sized for an LLM context instead of megabytes of base64.
Requirements
- Node.js 20+
- REW 5.40+ running with its API server enabled:
- macOS:
open -a REW.app --args -api - Windows:
roomeqwizard.exe -api - or enable it in REW Preferences → API
- macOS:
- API-triggered measurement (
run_sweep) additionally requires a REW Pro upgrade; every read/analysis/EQ tool works without it.
Install
// Claude Desktop (claude_desktop_config.json) or any MCP client config
{
"mcpServers": {
"rew": {
"command": "npx",
"args": ["-y", "room-eq-wizard-mcp"]
}
}
}
# Claude Code
claude mcp add rew -- npx -y room-eq-wizard-mcp
The REW API address defaults to http://127.0.0.1:4735; override with the REW_API_URL environment variable if you run REW on another port.
Tools (46)
| Area | Tools |
|---|---|
| Status | status, list_api_commands |
| Measurements | list_measurements, get_measurement, rename_measurement, delete_measurement, save_all_measurements, load_measurement_files, run_measurement_command, get_measurement_commands |
| Groups | list_groups, create_group, update_group, delete_group, add_measurements_to_group, get_group_measurements |
| Data | get_frequency_response, get_rt60, get_distortion |
| Measuring | run_sweep |
| Import | import_frequency_response, import_impulse_response, import_frequency_response_data, import_impulse_response_data, import_rta_file, import_sweep_recordings |
| Signal generator | generator |
| SPL meter | read_spl |
| EQ | auto_eq, get_eq_filters, set_eq_filters, get_predicted_response, list_equalisers |
| Processing | average_measurements, align_spl, arithmetic, smooth_measurement, add_spl_offset |
| Alignment | align_measurements, create_aligned_sum, get_alignment_state, configure_alignment, run_alignment_command |
| Analysis | analyze_response, compare_measurements, room_mode_analysis |
Typical session: status → run_sweep (or load_measurement_files) → analyze_response → room_mode_analysis with your room dimensions → auto_eq → compare_measurements on the predicted result.
Architecture
src/rew/ the only layer that talks to REW
codec.ts base64 big-endian float32 <-> arrays (REW's wire format)
types.ts Zod schemas — raw JSON is parsed once at this boundary
client.ts HTTP client; enables REW blocking mode once for long commands
src/analysis/ pure functions, no I/O, no mocks needed to test
spectrum.ts median-baseline peak/null detection, band stats, decimation
room-modes.ts rectangular-room modes, Schroeder frequency, correlation
src/tools/ MCP tools as data: one ToolDef type, one registration loop
Two properties worth knowing:
- Everything REW computes stays REW's job. RT60, averaging, EQ matching are invoked via the API, never reimplemented — the numbers you get through this server match what the REW GUI shows.
- Long commands use REW's blocking mode, enabled once per session. A sweep or EQ match returns when it finishes (client timeout 180 s). Note the blocking setting persists in REW across restarts.
Development
pnpm install
pnpm test # unit suite (mocked HTTP) + live suite that self-skips without REW
pnpm build
pnpm typecheck
With REW running (-api), the live integration tests in src/rew/live.integration.test.ts activate automatically.
Credits
Design informed by two earlier REW MCP servers: koltyj/REW-mcp (analysis depth, typed client patterns) and KevinMeinon/rew-mcp-server (API coverage philosophy and field-tested REW payload shapes, plus its excellent REW API reference doc). REW itself is © John Mulcahy.
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。