# Stunt Double

Runs AI personas through your web app's user journeys and reports where people get stuck, before real users hit the same wall.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI · ★ 1
- Author: stunt-double
- License: MIT · policy=link-only · description only; the file stays with the author
- Source: https://github.com/stunt-double/stuntdouble-mcp
- Updated: 2026-10-16
- HTML: https://affpapa.org/en/nevibecode/mcp/stunt-double

## What it does

Stunt Double is a hosted MCP server that plugs into your coding agent and exposes the Stunt Double platform. There you build actors — AI personas backed by knowledge entries — and push them through workflows and checklists on any web app. What comes back is a report on the friction points.

This repo ships no runnable server and no `package.json`. It holds plugin and MCP configuration, skills under `skills/`, agents under `agents/`, and Cursor rules. The endpoint itself is hosted at `https://app.stuntdouble.io/api/mcp`, so nothing works without reaching it. Skills install separately through the skills CLI and also need the server connected.

The skills split the work: `create-actor-panel` builds a persona, `run-ux-validation` runs a journey, `verify-change` walks an actor through the flows your latest change touched, and `maintain-automations` edits automations without losing run history. Separate skills audit brand and tone of voice, a live product against its design system, and legal requirements such as cookie consent and required disclosures. Some of them lean on the Stunt Double Index, which scores how well AI agents can find and act on a site.

## Requirements
- Remote server: `https://app.stuntdouble.io/api/mcp`

## Install

### Claude Code (native platform)

**Platform CLI: remote server**
```bash
claude mcp add --transport http mcp-server https://app.stuntdouble.io/api/mcp
```

### Codex

**Platform CLI: remote server**
```bash
codex mcp add stunt-double --url https://app.stuntdouble.io/api/mcp
```

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "stunt-double": {
            "url": "https://app.stuntdouble.io/api/mcp"
        }
    }
}
```

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "stunt-double": {
            "type": "http",
            "url": "https://app.stuntdouble.io/api/mcp"
        }
    }
}
```

### Gemini CLI

**Platform CLI: remote server**
```bash
gemini mcp add -t http stunt-double https://app.stuntdouble.io/api/mcp
```

## When to use it
- run a booking flow and see where users drop off
- check a fresh code change against the flows it affects
- build a panel of AI personas with product knowledge
- compare a live product with its design system
- gather evidence on cookie consent and required disclosures

## Example prompts
- Create an actor for the checkout journey and run it through the flow
- Verify my last change: send an actor through the affected user flows
- Audit the pricing page against our brand and tone-of-voice guidelines

## Similar
- [user-journeys](https://affpapa.org/en/nevibecode/skills/user-journeys.md) — skill · 581 installs
- [ui-ux-tester](https://affpapa.org/en/nevibecode/agents/ui-ux-tester.md) — subagent · ★ 26093
- [ai-first-engineering](https://affpapa.org/en/nevibecode/skills/ai-first-engineering.md) — skill · 4600 installs
- [agent-eval](https://affpapa.org/en/nevibecode/skills/agent-eval.md) — skill · 4500 installs
- [laravel-ai-sdk](https://affpapa.org/en/nevibecode/skills/laravel-ai-sdk.md) — skill · 667 installs
- [nielsen-heuristics-audit](https://affpapa.org/en/nevibecode/skills/nielsen-heuristics-audit.md) — skill · 739 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