quickcast-mcp
# QuickCast MCP (Screen-to-Action Protocol)
> **Autonomous AI Protocol for Screen Recordings**
> Turn screen recordings directly into structured GitHub bug reports with video timestamps, executable Playwright E2E tests, and Standard Operating Procedure (SOP) documentation.
Part of the **QuickCast Screen-Recorder** & **Watermark & Resize Studio** ecosystem.
---
## ⚡ Features
* **🎥 `get_latest_session`**: Instantly fetches recording metadata, duration, Cloudflare R2 watch URL, and user interaction markers from the latest session.
* **📋 `list_recent_sessions`**: Inspects recent screen recordings with durations, timestamps, and shareable preview links.
* **🐛 `format_github_bug_report`**: Formats an engineering-ready GitHub or Jira issue containing timestamped reproduction links (e.g. `[00:15](watchUrl#t=15)`), environment tables (URL, OS, resolution, browser), and failure descriptions.
* **🎭 `scaffold_playwright_test`**: Automatically outputs an executable Playwright reproduction test (in TypeScript or JavaScript) mirroring the exact interaction flow.
* **📖 `generate_sop_guide`**: Synthesizes a clean Markdown Standard Operating Procedure (SOP) for team training and workflow documentation.
* **📥 `ingest_session`**: Allows AI agents or scripts to register and store new recording telemetry into the local profile database (`~/.quickcast/sessions.json`).
---
## 📦 Installation & Setup
### 1. Claude Desktop Configuration
Add the following to your `claude_desktop_config.json`:
* **Windows**: `%APPDATA%\Claude\claude_desktop_config.json`
* **macOS**: `~/Library/Application Support/Claude/claude_desktop_config.json`
```json
{
"mcpServers": {
"quickcast": {
"command": "node",
"args": [
"C:\\Users\\Milmann\\.gemini\\antigravity\\scratch\\quickcast-mcp\\src\\index.js"
]
}
}
}
```
### 2. Cursor Configuration
Add to your project's `.cursor/mcp.json` or global Cursor MCP Settings:
```json
{
"mcpServers": {
"quickcast": {
"command": "node",
"args": [
"C:\\Users\\Milmann\\.gemini\\antigravity\\scratch\\quickcast-mcp\\src\\index.js"
]
}
}
}
```
---
## 🛠️ MCP Tools Reference
| Tool | Description | Inputs |
| :--- | :--- | :--- |
| `get_latest_session` | Get latest recording details & R2 watch URL | `session_id`, `watch_url`, `session_file` |
| `list_recent_sessions` | List last 10 recordings with durations & links | `limit`, `session_file` |
| `format_github_bug_report` | Generate markdown bug report with video timestamps | `issue_title`, `expected_behavior`, `actual_behavior` |
| `scaffold_playwright_test` | Generate executable Playwright E2E script | `language` (`typescript` / `javascript`), `test_name` |
| `generate_sop_guide` | Generate SOP documentation | `workflow_name`, `title` |
| `ingest_session` | Store session telemetry to local DB | `watchUrl`, `durationSeconds`, `markers`, `environment` |
---
## 🔒 Safety & Privacy
* **100% Client-Side & Local**: Runs as a local `stdio` server on your machine.
* **Zero Cloud Costs**: Uses your existing Cloudflare R2 links or local recording files. No external AI API keys or third-party cloud brokers required.
* **Decoupled Architecture**: Strictly isolated from extension recording engines and web apps.
---
## 📄 License
MIT © [Watermark & Resize Studio](https://watermarkresizestudio.com)
TDQS
Scored across 6 tools
Each tool has a distinct purpose: listing vs. retrieving session metadata, ingesting, and three different output-artifact generators (SOP, bug report, Playwright test). The only mild overlap is between get_latest_session and list_recent_sessions, but their descriptions clarify the distinction well.
All six tools follow a consistent snake_case verb_noun pattern (generate_sop_guide, list_recent_sessions, get_latest_session, format_github_bug_report, scaffold_playwright_test, ingest_session). The verbs vary appropriately to match each action without breaking the predictable convention.
Six tools is a well-scoped set for a session-recording processing server, covering retrieval, ingestion, and three distinct artifact generators without redundancy or padding.
The surface covers ingestion, listing, retrieval, and several generation workflows, which fits the domain well. Minor gaps remain, such as no delete/update or search-by-content operation, but agents can work around these.