Talkback MCP server
io.github.jmedure/talkback·v0.1.3·25 starsChat with your Ableton session in real-time.
25 stars53 downloads/wk
Reviews
Write oneNobody has reviewed Talkback yet.
If you have run it, two minutes of your experience saves the next person an afternoon.
Talkback tools (10, 2 write)
write = sends, deletes, buys or postsRead from the package source without running it. The installed server may list more.
analyze_mixwrite actionRuns rule-based heuristic analysis for potential mix issues: frequency buildup, dynamics problems, routing inefficiencies, headroom. Only call when the user explicitly asks for mix feedback or describes a specific problem — do NOT call proactively. Findings are suggestions to consider, not problems to fix. Mix quality is subjective — frame results in context of what the user is trying to achieve.
create_group_trackwrite actionCreates a new group track containing the specified tracks. Always confirm with the producer first.
get_bridge_healthReturns performance metrics from the M4L bridge device: poll execution time (avg/max), WebSocket message size, LiveAPI cache size, and track count. Use to diagnose performance issues or confirm the device is running efficiently.
get_plugin_libraryReturns all audio plugins (AU, VST3) installed on this system. Use to know what tools the producer has available.
get_session_contextReturns a snapshot of the current Ableton session: all tracks (names, volumes, panning, mutes, sends, device chains), return tracks, master track, and routing. Call once at the start of a conversation to orient yourself. No need to call again unless the user says the session has changed.
get_spectral_snapshotCaptures ~2 seconds of live audio from the master bus and returns peak/RMS levels across frequency bands. Transport MUST be playing — coordinate with the user on which section to analyze (e.g., 'play the chorus'). This is a point-in-time microscope, not a full-song analyzer. Use as a secondary check to verify what device parameter analysis suggests, or to detect issues invisible in the device chai
get_track_detailsReturns full detail for a specific track including all devices, every parameter in human-readable units, and per-device observations.
set_device_parameterSets a device parameter to a new value. Pass values in human-readable units (dB, ms, Hz, ratio, percent). NEVER call without explicit user approval. Present the planned change with current → proposed values first, wait for a clear 'yes', then execute. If the user modifies the plan, re-present and re-confirm before executing.
set_track_routingChanges a track's output routing. Always confirm with the producer first.
toggle_device_bypassEnables or bypasses a device on a track. Useful for A/B comparison. Always confirm with the user before toggling.
Public scan report
scanner v0.1.10 · 2026-09-27 · same rubric, same numbers if you re-run it
- Code scan35 source files scanned13/25
- –Live reliabilityno gateway calls yet and no remote to proben/a
- –Tool poisoningtools not inspected (local package is not executed); not countedn/a
- Auth qualitylocal package, no credentials required12/15
- Maintenancelast push 187 days ago4/15
- Maintainer identityregistry namespace matches repository owner; GitHub account older than a year8/10
Findings (1)
- highShell command built from a string (injection risk)
exec.shell-concatdist/ws-bridge.js: … try { const pids = execSync(`lsof -ti:${port}`, { encoding: "utf8" }).trim(); …
How Talkback compares
Against every graded MCP server, from our own scans and uptime checks. Updated after each scan run.
- Safety score
- 57/100
- higher than 12% of 48,255 servers
- Tools
- 10
- the typical server has 7
- Tools that change things
- 20%
- typical: 0%. Fewer write tools means less an agent can break
- GitHub stars
- 25
- more than 91% of 28,688 servers
- Downloads a week
- 53
- more than 44% of 12,191 servers with a package
What its tools let an agent do: 5 read · 1 create · 4 change
Install Talkback in Claude Code, Cursor or VS Code
Runs npx -y talkback-mcp on your machine. Read the scan report first; the gateway never runs local packages.
claude mcp add talkback -- npx -y talkback-mcp
What the publisher says
From the Talkback repository's README, as published. We do not edit it. Read it on GitHub
talkback mcp
Chat with your Ableton session in natural language. Go deeper in your mix, make it sound like it does in your head, and finally understand compression... jk no one can teach you that.
Talkback is an MCP server and Max for Live device that gives AI assistants real-time access to your Ableton Live session — read tracks, adjust parameters, analyze your mix, and get spectral snapshots, all through conversation.
talkback.createwcare.com · Docs · Changelog
How it works
Ableton Live ↔ M4L Bridge ↔ WebSocket ↔ MCP Server ↔ Claude / Cursor / etc.- talkback bridge — a Max for Live device on your master track that reads the Live Object Model and streams session data over WebSocket
- talkback-mcp (this package) — an MCP server that exposes your session as tools any LLM can use
Quick start
The fastest way to get set up — installs the MCP server and configures Claude Desktop automatically:
curl -fsSL https://talkback.createwcare.com/install.sh | bashAlready in an AI agent? Let it set things up for you:
npx skills add jmedure/talkback-mcpThen ask your agent: "Help me set up talkback"
Requires Node.js 18+. See the full setup guide for more options.
1. Install the Max for Live device
Download talkback-bridge (latest) and drop it onto your master track in Ableton. Make sure the device is toggled on.
2. Add the MCP server
Talkback requires a desktop MCP client — it runs locally on your machine and connects to Ableton over WebSocket. It does not work with claude.ai in the browser (web chat doesn't support local MCP servers yet).
Claude Desktop (recommended)
Download Claude Desktop, then add to your config:
~/Library/Application Support/Claude/claudedesktopconfig.json (macOS) %APPDATA%\Claude\claudedesktopconfig.json (Windows)
{
"mcpServers": {
"talkback-mcp": {
"command": "npx",
"args": ["-y", "talkback-mcp"],
"env": { "WS_PORT": "8765" }
}
}
}Restart Claude Desktop after saving.
Using nvm or fnm? Claude Desktop doesn't load your shell profile, so it may pick up an old system Node. Use the full path to npx instead — run which npx in Terminal, then set that as "command" in the config above. See troubleshooting for details.
Claude Code
claude mcp add --transport stdio talkback-mcp -- npx -y talkback-mcpCursor
Add to .cursor/mcp.json in your project root:
{
"mcpServers": {
"talkback-mcp": {
"command": "npx",
"args": ["-y", "talkback-mcp"],
"env": { "WS_PORT": "8765" }
}
}
}Other clients (Windsurf, etc.)
Point your client at npx -y talkback-mcp as the server command with WS_PORT=8765 as an environment variable.
Or use add-mcp to auto-configure all your installed agents:
npx add-mcp "npx -y talkback-mcp"3. Start chatting
Open your LLM and start asking about your session:
- "What's going on in my session?"
- "My bass sounds muddy, can you help?"
- "Cut 3 dB at 300 Hz on the vocal EQ"
- "Does my mix have any obvious problems?"
Your LLM will ask for approval before making any parameter changes. Undo always works.
Tools
Shortened. The full README is on GitHub.
Nothing above is checked by us. What we check is on the safety report.
Talkback: common questions
- Is Talkback MCP server safe?
- With care: it is graded C, so read the findings first (57/100). Read the Talkback safety report
- How do I install Talkback?
- It runs on your machine. Copy the Claude Code, Cursor, VS Code or Claude Desktop config from the install section.
- Does Talkback need an API key?
- No secret keys are declared. It reads 1 setting from the environment.
- Is Talkback maintained?
- The last commit was 225 days ago (2026-03-24). The latest release is v0.1.3.
- How does Talkback compare with other MCP servers?
- safety score 57/100 (higher than 12% of 48,255 servers); tools 10 (the typical server has 7); tools that change things 20% (typical: 0%. Fewer write tools means less an agent can break). Full comparison.