Hand the tedious parts of REAPER to your AI assistant.

Safety and privacy

Your assistant does the clicking. The decisions about your music stay with you.

• Nothing happens until you ask. Your assistant changes your project only when you ask it to.
• Every change is one undo step. Each change goes into REAPER's own undo history under the name of the tool that made it, so Ctrl+Z takes back one change at a time.
• It never saves unless you ask. Close the project without saving, and every change the assistant made is gone.
• It leaves your session where you had it. When your assistant adds audio, duplicates an item or measures, your track selection, time selection and edit cursor end up where you left them.
• Your audio stays on your computer. REAPER MCP talks to REAPER through files on your own machine and makes no network connections. Your assistant sees track names, settings and measurements, and those go to the AI model you use, the same as anything you type to it.
• It doesn't judge how your mix sounds. It can tell your assistant how loud a mix is, where its energy sits and how wide it is. Whether the vocal sounds harsh is still your call.

Before a big AI-driven change, save a copy first or lean on REAPER's project auto-backups, the same as you would before any major edit.

Supported AI apps

REAPER MCP connects through MCP, the standard that AI apps use to work with other software. It works with Claude Desktop, the ChatGPT desktop app and Codex, Gemini CLI, GitHub Copilot, Cursor and other apps that run local MCP servers (full list).

It's free and open source under the MIT license. It needs no API key and no subscription of its own, because it works inside the AI app you already have. It runs on Windows, macOS and Linux.

Testing and support

The changelog, the release history and the automated tests are public in the REAPER MCP repository.

How it works

REAPER can't take requests from other programs on its own, so REAPER MCP has two parts. Your AI app runs the MCP server, which turns your requests into REAPER actions. A small bridge script runs inside REAPER and carries those actions out in your open project, so keep REAPER open with the project you want to work on.

Setup

Requirements

• REAPER (any recent version; live-tested through 7.79) on Windows, macOS, or Linux
• Python 3.10 or newer, handled for you by uv or pipx in the steps below
• An AI app that can run a local MCP server, such as Claude Desktop, the ChatGPT desktop app or Codex, Gemini CLI, GitHub Copilot or Cursor (full list)

Setup with your AI assistant

If your app has an agent that can run terminal commands, paste this in. It handles the install and the settings, then walks you through the one step REAPER needs from you. If your app can't run commands, follow the steps below instead.

Set up the TwelveTake REAPER MCP server for me.
- It's a Python package on PyPI called twelvetake-reaper-mcp. Install it so it can
  run (uvx or pipx) and add it to my MCP client config as a server named "reaper".
- REAPER needs a small bridge script. Run "uvx twelvetake-reaper-mcp --install-bridge"
  to copy it into my REAPER Scripts folder. Ask me first whether REAPER should start
  the bridge by itself at launch, and if I say yes, add --autostart to that command.
- Then tell me the one manual step: in REAPER, open Actions -> Show action list ->
  Load ReaScript, choose reaper_mcp_bridge.lua, and Run it. I should see a
  "REAPER MCP Bridge ... started" message in REAPER's console.
- Confirm it works by asking how many tracks are in my project.

Manual setup

Step 1: Install the bridge in REAPER

One command installs the bridge script and the small analyzer plugin that the measurement tools use, each in the folder REAPER expects, on any platform. If it finds a different copy of the bridge, it backs that copy up first. It writes nothing else:

uvx twelvetake-reaper-mcp --install-bridge

Then load it in REAPER:

1. 1. Open Actions → Show action list.

   

2. 2. Click New action → Load ReaScript.

   

3. 3. Select reaper_mcp_bridge.lua in the Scripts folder and click Open.

   

4. 4. With the script selected in the action list, click Run.

   

5. 5. REAPER's console confirms the bridge is running.

   

Next time, type mcp bridge in the action list's filter box to find it. To have REAPER start the bridge by itself every time it launches, add --autostart:

uvx twelvetake-reaper-mcp --install-bridge --autostart

That adds a marked block to __startup.lua in REAPER's Scripts folder, backing the file up first, and leaves anything else in it alone. --remove-autostart takes the block out again.

Use ReaPack? Install the bridge from there instead. Open Extensions → ReaPack → Import repositories and paste this address:

https://github.com/TwelveTake-Studios/reaper-mcp/raw/main/index.xml

Then open Extensions → ReaPack → Browse packages, find TwelveTake REAPER MCP bridge, right-click it, choose Install, and click OK:

ReaPack adds the script to the action list and installs the analyzer. Run the script from there as in steps 4 and 5 above. To have REAPER start the ReaPack copy at launch, run uvx twelvetake-reaper-mcp --autostart.

To place it by hand instead, download reaper_mcp_bridge.lua from the REAPER MCP repository, copy it into your REAPER Scripts folder, and load it as above. For the measurement tools, also copy twelvetake_mcp_analyzer.jsfx into Effects/TwelveTake, next to the Scripts folder:

Windows: %APPDATA%\REAPER\Scripts\
macOS:   ~/Library/Application Support/REAPER/Scripts/
Linux:   ~/.config/REAPER/Scripts/

Step 2: Add the server to your AI app

Nearly every AI app uses the same JSON for this: an mcpServers entry with a command and its args. Add this to your app's MCP settings:

{
  "mcpServers": {
    "reaper": {
      "command": "uvx",
      "args": ["twelvetake-reaper-mcp"]
    }
  }
}

Where you put it depends on your app:

• Claude Desktop: Settings → Developer → Edit Config opens claude_desktop_config.json (in %APPDATA%\Claude\ on Windows, ~/Library/Application Support/Claude/ on macOS). Paste the block above, save, then fully quit and reopen Claude.
• Gemini CLI: add the block above to ~/.gemini/settings.json: same shape.
• Cursor: Settings → MCP → Add, or paste the block above into ~/.cursor/mcp.json: same shape.
• Devin Desktop (formerly Windsurf): edit ~/.config/devin/mcp_config.json (%AppData%\devin\mcp_config.json on Windows): same shape.
• VS Code: Command Palette → "MCP: Add Server," or edit .vscode/mcp.json. VS Code uses a top-level servers key with "type": "stdio" instead of mcpServers.

The ChatGPT desktop app and Codex share one settings file, ~/.codex/config.toml, which uses a different format. With Codex CLI installed, run codex mcp add reaper -- uvx twelvetake-reaper-mcp, or add this to the file:

[mcp_servers.reaper]
command = "uvx"
args = ["twelvetake-reaper-mcp"]

Using a different app? Add the same command and args wherever it lists MCP servers. After adding it, fully quit and reopen the app.

Step 3: Check it's working

With REAPER open and the bridge running, ask your assistant: "How many tracks are in my project?" A number back means everything's connected.

Updates

uvx keeps using the version it first downloaded, so updating takes one command. It fetches the newest release and copies its bridge into REAPER's Scripts folder, and the running bridge reloads itself. You don't have to do anything in REAPER:

uvx twelvetake-reaper-mcp@latest --install-bridge

If you installed with pipx, run pipx upgrade twelvetake-reaper-mcp and then twelvetake-reaper-mcp --install-bridge. With ReaPack, Extensions → ReaPack → Synchronize packages fetches the new bridge. It runs the next time the bridge starts, or right away if you run uvx twelvetake-reaper-mcp --reload-bridge.

Either way, restart your AI app so it picks up the new server too. If you're updating from 1.7.7 or older, run the bridge in REAPER once more after installing it. From 1.7.8 on, it reloads by itself.

Troubleshooting

• No "Bridge … started" message? Run reaper_mcp_bridge.lua from Actions → Show action list, and check that the file is in your Scripts folder.
• "Not started: another copy is already running"? The bridge was already running, so you don't need to do anything. Only one copy runs at a time.
• Your assistant can't see the "reaper" server? Fully quit and reopen your app after editing its settings, and check that the JSON is valid (a stray trailing comma is the usual culprit).
• "Cannot connect to REAPER"? Make sure REAPER is open and the bridge is running this session. It stops when REAPER closes, unless you set it to start with REAPER (--autostart).

Feature list

You don't call these yourself. You ask in plain language, and your assistant picks the right ones.

Limitations

• The mastering chain is a starting point. It inserts ReaEQ, ReaComp, ReaEQ and ReaLimit on the master track at their default settings, and you or your assistant set them from there.
• Third-party plugins take raw values. Your assistant can read and set any plugin parameter, but outside ReaEQ's bands the values are on REAPER's 0 to 1 scale rather than in dB or Hz.

FAQ