Voice automation, documented.

The CLI is a signed helper that ships inside the app. One click links it into your shell.

export PATH="$HOME/.local/bin:$PATH"

murmur status

Pass text directly, or point Murmur at a UTF-8 text or Markdown script. Output can be WAV or M4A.

murmur generate \
  --text "Welcome back. Today we're building a local voice workflow." \
  --voice af_bella \
  --speed 0.95 \
  --output audio/intro.wav

murmur generate \
  --input scripts/chapter-01.txt \
  --model kokoro \
  --output audio/chapter-01.m4a

MCP is the agent-facing layer over the same CLI contract. Each setup uses the helper’s absolute path, so it works without any shell PATH configuration.

codex mcp add murmur -- /Applications/Murmur.app/Contents/Helpers/murmur mcp serve

claude mcp add --transport stdio --scope user murmur -- /Applications/Murmur.app/Contents/Helpers/murmur mcp serve

{
  "mcpServers": {
    "murmur": {
      "command": "/Applications/Murmur.app/Contents/Helpers/murmur",
      "args": ["mcp", "serve"]
    }
  }
}

The agent can inspect, download, select, and remove models; inspect voices; then submit audio work and return the exact files it created. Long operations return a job ID by default, so the agent can keep working and poll murmur_get_job without blocking the MCP connection.

Keep a small JSON manifest beside the video script. Relative input and output paths resolve from the manifest’s folder, so the voiceover versions with the project.

{
  "items": [
    {
      "id": "intro",
      "text": "Here is the feature we are building today.",
      "outputPath": "audio/01-intro.wav",
      "voiceID": "af_bella",
      "speed": 0.95
    },
    {
      "id": "demo",
      "inputPath": "scripts/02-demo.txt",
      "outputPath": "audio/02-demo.wav",
      "voiceID": "af_bella"
    }
  ]
}

murmur batch --manifest video/scenes.json

CLI reference

MCP tools