Skip to main content
Glama
REW MCP Server

REW MCP Server

by manuaero

REW MCP Server

CI Python 3.12+ Server code: MIT

Control Room EQ Wizard from Codex, Claude Desktop, or another MCP client. Measure, inspect responses, work with EQ and alignment, configure audio and simulation, and export data through REW's local HTTP API.

The bridge discovers REW's API instead of maintaining a fixed list of wrappers. The included REW 5.40 Beta 148 / API 0.9.9 snapshot contains 608 operations across 384 paths. All of those operations can be discovered and called through 16 compact MCP tools. Optional --all-tools mode adds an individual tool for each operation: 624 tools for that snapshot. Live coverage follows the connected REW version.

flowchart LR
    Client[Codex / Claude / MCP client] <-->|MCP over stdio| Bridge[REW MCP Server]
    Bridge <-->|HTTP on loopback| REW[Room EQ Wizard 5.40+]
    REW -->|doc.json| Catalog[Live API catalog]
    Catalog --> Bridge

What it can do

Area

Capabilities exposed by REW

Measurements

Sweeps, summaries, UUID selection, frequency/impulse responses, group delay, distortion, RT60, windows, smoothing, arithmetic, and groups

EQ and alignment

Targets, house curves, filters, equaliser models, optimization, predicted responses, phase/delay alignment, and aligned sums

Audio and generator

Java/ASIO routing, channels, sample rates, input levels, signal configuration, level, frequency, and play/stop

Live analysis

SPL meters/logger, RTA, stepped measurements, and available progress endpoints

Room simulator

Dimensions, surfaces, sources, positions, configuration, and responses

Files and application

MDAT save/load, response imports, supported REW export commands, graphs, image capture, warnings, and errors

The bridge also decodes REW's Base64 float arrays, reconstructs frequency/time coordinates, calculates response statistics, exports CSV/JSON, writes house curves, and paginates large results without silently truncating them.

Requirements and limits: Python 3.12+, REW 5.40+ with its API enabled, and an MCP client with stdio support. Automated sweep measurements require REW Pro. GUI-only features are accessible only where REW provides an API operation. Callbacks require a receiver supplied by the integrating application. The bridge runs on the same computer as REW and does not start audio at startup.

Related MCP server: sonora-mcp

Quick start

Clone the repository:

git clone https://github.com/manuaero/rew-mcp-server.git
cd rew-mcp-server

Windows

Install Python 3.12+ and an API-enabled REW version from the official REW downloads. Then run in PowerShell:

.\scripts\Setup.ps1 -RegisterCodex
.\scripts\Start-REW.ps1
.\.venv\Scripts\python.exe -m rew_mcp.server --check

-RegisterCodex is optional; it requires the Codex CLI. Supply -Python C:\path\to\python.exe or -RewPath C:\path\to\roomeqwizard.exe when using non-default installations. The setup script installs this bridge; REW is installed separately.

macOS and Linux

python3 -m venv .venv
.venv/bin/python -m pip install -e '.[dev]'

Start REW with its API enabled. On macOS:

open -a REW --args -api

On Linux, launch your REW executable with -api. Verify the connection:

.venv/bin/python -m rew_mcp.server --check

Connect your MCP client

For Codex, use the absolute path to the virtual environment's Python:

codex mcp add rew -- /absolute/path/rew-mcp-server/.venv/bin/python -m rew_mcp.server

For Windows, the interpreter is .venv\Scripts\python.exe. For Claude Desktop and other clients, start with examples/mcp-config.json and replace the example paths. Restart the MCP connection after registration.

See installation and client configuration for complete instructions.

Try it

  • "Check the REW connection and list my measurements."

  • "Analyze this measurement between 20 and 200 Hz."

  • "Read the current audio routing and sweep settings."

  • "Find the available equaliser models and target settings."

  • "Inspect the alignment tool operations for my main and subwoofer."

  • "Export this frequency response as CSV."

The normal workflow is search → describe → call:

{"name":"rew_search_api","arguments":{"search":"target","tag":"eq"}}

Use the returned operation ID with rew_describe_api to inspect its exact parameters, then call rew_call_api. rew_request also accepts a documented method/path directly. Use measurement UUIDs because measurement indices change.

Before playback, inspect actual routing, levels, and protection settings. A successful HTTP 202 means accepted, not completed. The bridge never automatically retries a mutation after a timeout.

Documentation

Guide

Contents

Installation

Windows/macOS/Linux setup, REW startup, headless testing, upgrades, and removal

Client configuration

Codex, Claude Desktop, other stdio clients, paths, timeouts, and environment variables

Tool reference

All 16 MCP tools and their generated input schemas

API catalog

Every operation in the bundled REW snapshot, organized by category

Workflows

Response analysis, EQ, alignment, import/export, batching, and progress handling

Architecture

API discovery, validation, transport, data encoding, lifecycle, and project layout

Troubleshooting

Connection, licensing, parameter, async, file, and audio issues

Contributing

Development, tests, documentation generation, and pull requests

Security model

Local trust boundary, data handling, and reporting issues

Changelog

Version history

Development and tests

python -m pip install -e '.[dev]'
python -m pytest -q
python -m ruff check src tests scripts
python -m ruff format --check src tests scripts
python scripts/generate_docs.py --check

The default tests do not need REW. They cover the full bundled API catalog, validation, array decoding, analysis, offline behavior, timeouts, stale progress, file handling, workflows, and actual MCP stdio sessions in both tool modes.

The opt-in live smoke test uses a synthetic measurement and plays no audio:

python scripts/live_smoke.py

It temporarily adds and removes its own measurement. It does not verify physical audio I/O, Pro-only sweeps, or every mutation. CI runs the offline tests and builds the package; it does not run the live test.

Acknowledgments and license

Inspired by KevinMeinon/rew-mcp-server. This is an independent implementation using the official MCP Python SDK. Room EQ Wizard is developed by John Mulcahy; see the REW API documentation.

The server code is MIT licensed. The bundled REW API specification retains its source metadata and REW EULA reference; it is not relicensed as part of this server. This project does not include REW binaries and is not an official REW product.

Available Tools

16 tools
rew_analyze_measurementB
Read-only

Calculate response statistics over a frequency range. Reports observed extrema without inferring acoustic causes or inventing EQ.

Inputs: measurement_id (required), query, max_hz, min_hz.

rew_batchB
Destructive

Execute an ordered API workflow. Stops at first error or uncompleted asynchronous action. Earlier actions are not rolled back. Use UUIDs, not indices.

Inputs: steps (required).

rew_call_apiB
Destructive

Execute any discovered REW operation with validated arguments. Body supports objects, arrays and JSON scalars. Waiting requires a documented progress endpoint. Timeout does not mean the action was cancelled.

Inputs: operation_id (required), body, wait, query, +3 more.

rew_commandB
Destructive

Run a REW command at a /command endpoint. Discover command choices and parameters first. Handles generator, sweeps (Pro), EQ, alignment, RTA, measurement processing, save/load and other command scopes.

Inputs: path (required), command (required), wait, parameters, +3 more.

rew_decode_arrayB
Read-only

Decode REW big-endian IEEE-754 float32 Base64 samples. Non-finite values are represented as null.

Inputs: encoded (required), offset, max_points.

rew_describe_apiA
Read-only

Read a discovered operation's exact parameter schemas, request body, units and responses before calling it.

Inputs: operation_id (required).

rew_encode_arrayB
Read-only

Encode numeric samples for REW frequency/impulse data imports.

Inputs: values (required).

rew_export_measurementC

Save complete decoded response data as CSV or JSON inside REW_OUTPUT_DIR. Existing files are not overwritten.

Inputs: filename (required), measurement_id (required), query, format, +1 more.

rew_get_measurementB
Read-only

Get measurement summary or endpoint data such as frequency-response, impulse-response, group-delay, distortion or rt60. Use UUID. Decode Base64 float32 data with frequency/time coordinates and paginated samples.

Inputs: measurement_id (required), query, decode, offset, +2 more.

rew_list_measurementsA
Read-only

List all loaded measurements, UUIDs, names and metadata.

Inputs: no named parameters.

rew_read_outputB
Read-only

Read a saved JSON result inside REW_OUTPUT_DIR using an optional JSON Pointer, paginating objects, arrays or long strings.

Inputs: filename (required), limit, offset, pointer.

rew_requestB
Destructive

Call any documented REW method/path, including operations without dedicated tools. Use rew_search_api and rew_describe_api to get exact schemas.

Inputs: path (required), method (required), body, wait, +3 more.

rew_search_apiB
Read-only

Search ALL REW operations by words, category tag or HTTP method. Covers audio devices, measurements, EQ, alignment, generator, SPL, RTA, stepped measurements, import/export, simulation, graphs and application settings.

Inputs: tag, limit, method, offset, +2 more.

rew_statusA
Read-only

Check live REW connection, API coverage and operating mode.

Inputs: no named parameters.

rew_stop_audioA

Attempt to stop generator, sweep, stepped measurement and RTA independently, bypassing this server's command wait queue.

Inputs: no named parameters.

rew_write_house_curveB

Create a REW house curve text file from [Hz, dB] points inside REW_OUTPUT_DIR. Then load through the discovered house curve endpoint.

Inputs: points (required), filename (required).

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    B3.3/5.0

    Scored across 16 tools

    Disambiguation3/5

    The measurement-specific, array codec, export/output and status tools are distinct. However rew_call_api, rew_request and rew_command all execute REW operations with subtle boundaries, and rew_batch also overlaps by running API workflows. Descriptions help, but an agent may still hesitate among the generic execution tools.

    Naming Consistency4/5

    All tools share the rew_ prefix and snake_case formatting, with mostly verb_noun forms such as rew_list_measurements and rew_get_measurement. A few names are bare nouns or actions like rew_status, rew_request, rew_command and rew_batch, so the pattern is mostly consistent rather than perfect.

    Tool Count4/5

    Sixteen tools is slightly above the ideal 3–15 range, but it is reasonable for a broad REW surface needing discovery, execution, measurement helpers and output handling. No tool appears wholly redundant, though the generic execution cluster adds some weight.

    Completeness4/5

    Measurement lifecycle is well covered through list/get/analyze/export, and API discovery plus generic execution can reach most REW operations. Generic call/request/command fill most gaps, though there is no explicit delete/rename/import tool and workflows rely on the generic surface.

    Maintenance

    ActivityMaintained
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      D
      maintenance
      Dynamically generates MCP tools from OpenAPI specifications, enabling AI assistants to interact with any REST API through natural language. Supports multiple APIs with authentication, parameter validation, and integration with Claude Desktop and LangChain.
      1
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      MCP server that wraps the Multiroom Audio Hub v2 REST API, providing 24 tools to control multiroom audio systems. Enables AI assistants to manage inputs, outputs, groups, routes, playback, and master mute.
      AGPL 3.0
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables voice assistants to start, monitor, message, and approve local Claude Code sessions on your computer, including sessions already open in a terminal. It exposes session management, recaps, approvals, and history search as MCP tools.
      7
      GPL 3.0
    • A
      license
      Not graded
      quality
      A
      maintenance
      Enables MCP-compatible clients to manage, run, and interact with local or cloud LLM models through exposed tools for listing, pulling, generating, chatting, embeddings, and web search/fetch. Supports Claude Desktop, Cline, and hybrid local/cloud setups.
      812 npm
      AGPL 3.0