Lost Media Search Plugin
A Claude Code plugin for tracking down lost, obscure, and forgotten media — TV shows, films, music, video games, commercials, broadcasts, and other vanished content.
Built on the methodology from The Lost Media and Research Handbook by Ziggy Cashmere, adapted for use with Claude Code's digital tools.
Installation
From the Marketplace
-
Ensure Claude Code has read access to the repo (public repos work automatically; for private repos, set up a GitHub token with repo access)
-
Add the marketplace:
/plugin marketplace add lawriec/lost-media-search-plugin
-
Install the plugin:
/plugin install lost-media-search@lost-media-search-plugin
-
Configure credentials. The plugin needs API keys and paths to wire up its MCP servers — see Configuring the Plugin below. You can either set shell environment variables before launching claude, or walk through the /plugin configure UI after install. Until you do one of these, the optional-MCP servers will fail to connect and the skill will fall back to Claude Code's built-in search.
Direct from GitHub
/plugin install lost-media-search@github:lawriec/lost-media-search-plugin
Same post-install step applies — see Configuring the Plugin.
What it does
This plugin turns Claude into a lost media researcher. It can:
- Track down media you remember but can't find
- Search archives for deleted or vanished content
- Identify unknown media from vague memories
- Research availability of obscure content
- Find old commercials, bumpers, and broadcasts
- Analyze audio/video files to determine origin and era
MCP Servers
The plugin configures 9 MCP servers for research tools:
| Server | Package | Purpose | API Key Required |
|---|
| searxng | github:lawriec/mcp-searxng | Primary search tool. Multi-engine web search via self-hosted SearXNG — queries Google, Bing, Brave, Yahoo, DuckDuckGo, and 250+ engines simultaneously. Full operator pass-through, category-based search (videos, files/torrents, science, music, social media), language filtering, and engine attribution. | No API key — requires Docker (see SearXNG Setup) |
| tavily | tavily-mcp@latest | Targeted web search with date filtering, content extraction, site crawling | Yes — TAVILY_API_KEY |
| gemini | github:lawriec/mcp-gemini-media | AI analysis of video and audio files | Yes — GEMINI_API_KEY |
| yt-dl | @kevinwatt/yt-dlp-mcp@latest | YouTube search, metadata, downloads | No (but cookies strongly recommended — see below) |
| internet-archive | github:lawriec/mcp-internet-archive | Internet Archive search and downloads | No |
| video-reader | github:lawriec/mcp-video-reader | Video analysis, thumbnails, ffmpeg operations | No |
| common-crawl | github:lawriec/mcp-common-crawl | Search and extract content from Common Crawl web archives | No |
| google-reverse-image | github:lawriec/mcp-google-reverse-image | Reverse image search via Google Cloud Vision Web Detection — finds pages containing an image, exact/partial/visually-similar matches, best-guess label, and detected web entities. Accepts local file paths or public URLs. | Yes — GOOGLE_VISION_API_KEY |
SearXNG Setup
SearXNG is the plugin's primary search tool. It requires a local Docker instance — without
it, searches fall back to the less capable WebSearch tool.
1. Install Docker if you don't have it: https://docs.docker.com/get-docker/
2. Start SearXNG using the pre-configured Docker Compose setup in the
mcp-searxng repo:
git clone https://github.com/lawriec/mcp-searxng.git
cd mcp-searxng/docker
docker compose up -d
3. Verify it's running:
curl "http://localhost:8080/search?q=test&format=json" | head -c 200
4. Note the URL — you'll be asked for it when you install the plugin (step below).
The default is http://localhost:8080.
SearXNG runs as a lightweight Docker container. Start it before launching Claude Code and
leave it running — it uses minimal resources when idle.
Optional: VPN protection — SearXNG queries upstream engines (Google, Bing, etc.) directly
from your IP. During intensive research sessions this can lead to rate-limiting or blocking.
The mcp-searxng Docker setup includes an optional OpenVPN sidecar that routes all search
traffic through a VPN tunnel:
- Place
.ovpn files from your VPN provider (NordVPN, ExpressVPN, Surfshark, ProtonVPN,
Mullvad, etc.) into a folder
- If your provider requires auth, add a
default.auth file (username line 1, password line 2)
- Set environment variables and start with the VPN override: