frevana

Frevana - Browser AI, Web Scraping & Social Publishing

Use the Frevana MCP tools to control Chrome: ask AI platforms questions, extract any web page as clean Markdown, and publish posts to social media.

Prerequisites

Before using Frevana tools, verify the connection:

mcp__frevana__frevana_status({})

If this fails, the MCP server is not running. Tell the user: "Please start the Frevana MCP server first — run /mcp and start frevana."

Available MCP Tools

• mcp__frevana__frevana_ask - Send a prompt to an AI platform and get the response
• mcp__frevana__frevana_scrape - Scrape content from any web page (uses Readability.js for clean extraction)
• mcp__frevana__frevana_publish - Publish a post to Twitter/X, Facebook, or LinkedIn
• mcp__frevana__frevana_status - Check connection status

Usage

Ask an AI platform a question

mcp__frevana__frevana_ask({
  provider: "chatgpt",  // chatgpt, gemini, perplexity, deepseek, doubao
  prompt: "your question here",
  timeout: 120000  // optional, default 2 minutes
})

Scrape any web page as Markdown

Uses Readability.js + Turndown to extract clean article content from any URL. Reuses your Chrome login sessions - can access paywalled/authenticated content.

mcp__frevana__frevana_scrape({
  url: "https://example.com/article",
  provider: "url",           // use "url" for universal extraction
  timeout: 60000             // optional
})

Returns: clean Markdown content, page title, author, extracted links, and metadata. Handles login detection (returns error if page requires login). Large pages are automatically truncated at 800KB.

Batch scrape multiple URLs

mcp__frevana__frevana_scrape({
  action: "SCRAPE_BATCH",
  urls: ["https://url1.com", "https://url2.com", "https://url3.com"],
  options: { concurrency: 3 }  // up to 3 tabs in parallel
})

Publish to social platforms

// Twitter/X
mcp__frevana__frevana_publish({
  provider: "twitter",
  text: "Your tweet content"
})

// Facebook
mcp__frevana__frevana_publish({
  provider: "facebook",
  text: "Your post content"
})

// LinkedIn
mcp__frevana__frevana_publish({
  provider: "linkedin",
  text: "<p>Your post content</p>"  // supports HTML for LinkedIn
})

Check status

mcp__frevana__frevana_status({})

Instructions

1. Always check status first if you get connection errors - call frevana_status
2. When the user says /frevana followed by a question, use frevana_ask with the appropriate provider
3. Default provider is chatgpt unless the user specifies otherwise
4. If the user mentions a specific AI (e.g., "ask Gemini", "use DeepSeek"), use that provider
5. For web scraping requests, use frevana_scrape with provider: "url" for any arbitrary URL
6. For social media publishing, use frevana_publish with the appropriate provider (twitter, facebook, linkedin)
7. Present results clearly - format Markdown content, note which provider answered
8. For batch scraping, use SCRAPE_BATCH to process multiple URLs in parallel
9. When user says /frevana help, show available features in English

Examples

User: /frevana ask ChatGPT to explain quantum computing → Call frevana_ask with provider="chatgpt", prompt="explain quantum computing"

User: /frevana ask deepseek to write a sorting algorithm → Call frevana_ask with provider="deepseek", prompt="write a sorting algorithm"

User: /frevana scrape https://news.ycombinator.com → Call frevana_scrape with provider="url", url="https://news.ycombinator.com"

User: /frevana scrape these 3 links → Call SCRAPE_BATCH with the 3 URLs, concurrency=3

User: /frevana tweet "Just shipped a new feature!" → Call frevana_publish with provider="twitter", text="Just shipped a new feature!"

User: /frevana post to LinkedIn "We just launched v2.0" → Call frevana_publish with provider="linkedin", text="We just launched v2.0"

User: /frevana compare ChatGPT and Gemini on "what is AGI" → Call frevana_ask twice in parallel: once with provider="chatgpt" and once with provider="gemini", same prompt