# Woodstitch

The Woodstitch MCP server for Claude Code sends text-to-speech audio to paid human listeners for blind evaluation. An agent can compare two clips or ask listeners to classify one as human-made or AI-generated.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI
- Author: woodstitch-llc
- License: MIT · policy=link-only · description only; the file stays with the author
- Source: https://github.com/woodstitch-llc/woodstitch-mcp
- Updated: 2026-11-07
- Security check: automated, 2026-11-17 (text) — no signs of malicious code found.
- HTML: https://affpapa.org/en/nevibecode/mcp/woodstitch

## What it does

Woodstitch submits audio for blind review by paid listeners and returns scores with confidence intervals. It supports two study formats: listeners choose between two clips of the same text, or classify a single clip as human or AI speech. MOS, CMOS, MUSHRA and 1–5 rating scales are outside its scope.

The server accepts uploaded local audio or public URLs, prepares a study quote and starts the study after the price is confirmed. It needs a Woodstitch API key, Python 3.10+ and `uvx`; creating a study can price it without charging, while starting it spends money. A confirmed amount caps the charge, and a higher quote prevents the study from starting.

## Requirements
- Needs: Python and uv (commands run via uvx)
- Environment variables: `WOODSTITCH_API_KEY` — keys and tokens: ask the human for the values, never commit them or put them in shared files.

## Install

### Claude Code (native platform)

**Platform CLI**
```bash
claude mcp add tts-human-eval -e WOODSTITCH_API_KEY=<WOODSTITCH_API_KEY> -- uvx woodstitch-mcp
```
Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from woodstitch-mcp <command>

### Codex

**Platform CLI**
```bash
codex mcp add woodstitch --env WOODSTITCH_API_KEY=<WOODSTITCH_API_KEY> -- uvx woodstitch-mcp
```
Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from woodstitch-mcp <command>

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "woodstitch": {
            "command": "uvx",
            "args": [
                "woodstitch-mcp"
            ],
            "env": {
                "WOODSTITCH_API_KEY": "<WOODSTITCH_API_KEY>"
            }
        }
    }
}
```
Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from woodstitch-mcp <command>

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "woodstitch": {
            "type": "stdio",
            "command": "uvx",
            "args": [
                "woodstitch-mcp"
            ],
            "env": {
                "WOODSTITCH_API_KEY": "<WOODSTITCH_API_KEY>"
            }
        }
    }
}
```
Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from woodstitch-mcp <command>

### Gemini CLI

**Platform CLI**
```bash
gemini mcp add -e WOODSTITCH_API_KEY=<WOODSTITCH_API_KEY> woodstitch uvx woodstitch-mcp
```
Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from woodstitch-mcp <command>

## When to use it
- compare two speech-synthesis versions of the same passage
- check whether a clip sounds human or AI-generated
- get a study quote before spending money
- export listener ratings as JSON or CSV

## Example prompts
- Compare the spoken quality of `base.wav` and `candidate.wav`.
- Classify `sample.wav` as human speech or synthesized speech.
- Quote a listening study for `a/001.wav` and `b/001.wav` without starting it.

## Similar
- [qianwen-audio-tts](https://affpapa.org/en/nevibecode/skills/qianwen-audio-tts.md) — skill · 6200 installs
- [sherpa-onnx-tts](https://affpapa.org/en/nevibecode/skills/sherpa-onnx-tts.md) — skill · 5400 installs
- [characteristic-voice](https://affpapa.org/en/nevibecode/skills/characteristic-voice.md) — skill · 2800 installs
- [qwencloud-audio-tts](https://affpapa.org/en/nevibecode/skills/qwencloud-audio-tts.md) — skill · 1800 installs
- [audio-router](https://affpapa.org/en/nevibecode/skills/audio-router.md) — skill · 1600 installs
- [stepfun-tts](https://affpapa.org/en/nevibecode/skills/stepfun-tts.md) — skill · 702 installs

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md
