Recast MCP
README.md
# Recast MCP
Turn any URL into platform-ready social media content. LinkedIn posts, Twitter threads, Reddit posts, and newsletter snippets — generated by your AI assistant.
## How it works
Recast MCP gives your AI assistant (Claude, Cursor, etc.) two capabilities:
1. **`fetch_content` tool** — Fetches any URL and extracts clean, readable text (strips ads, navigation, boilerplate)
2. **4 prompt templates** — Platform-specific content generation for LinkedIn, Twitter, Reddit, and Newsletter
Your AI does the writing. Recast handles the extraction and formatting rules.
## Install
```bash
npx recast-mcp
```
### Claude Desktop config
Add to `claude_desktop_config.json`:
```json
{
"mcpServers": {
"recast": {
"command": "npx",
"args": ["-y", "recast-mcp"]
}
}
}
```
### Cursor / VS Code
Add to your MCP settings:
```json
{
"recast": {
"command": "npx",
"args": ["-y", "recast-mcp"]
}
}
```
## Usage
Once installed, just tell your AI:
> "Fetch https://example.com/article and turn it into a LinkedIn post"
Or use the tools directly:
1. Call `fetch_content` with a URL
2. Use any of the 4 prompt templates with the fetched content:
- `repurpose_linkedin` — Professional post with hook, CTA, and hashtags
- `repurpose_twitter` — Thread of 3-7 tweets, each under 280 chars
- `repurpose_reddit` — Authentic post with TLDR and discussion question
- `repurpose_newsletter` — Personal snippet with insight on why it matters
### Tone options
Each prompt accepts a `tone` parameter:
- `professional` (default) — Polished, business-ready
- `casual` — Relaxed, conversational
- `technical` — Detailed, assumes domain expertise
## Supported sources
Works with any public URL. Automatically detects source type:
- Articles and blog posts
- YouTube pages (extracts description/metadata)
- Reddit threads
- Medium and Substack posts
- News articles
## Privacy
100% local. Content is fetched directly by your machine. No data is sent to any third-party service. Your AI assistant generates the content — no external API calls.
## License
MIT