MCP · Fetch — a URL becomes markdown context

Use case: pull a web page's content (docs, article, changelog, competitor pricing) into the answer — converted to clean markdown, more targeted than web search.

Configuration (.znote/mcp.json)

{
  "mcpServers": {
    "fetch": {
      "command": "uvx",
      "args": ["mcp-server-fetch"]
    }
  }
}

Python reference server, launched by uvx (install uv). It paginates long pages automatically (start_index).

Exposed tool: fetch(url, max_length, start_index, raw).

Examples

Summarize an article

Fetch https://example.com/blog/important-post and give me a 5-bullet
summary with the key quotes.

Competitive watch

Compare the pricing pages at https://competitor-a.com/pricing and
https://competitor-b.com/pricing: table of plans, prices, and
positioning differences.

Track a changelog

Read https://github.com/electron/electron/releases and tell me what
changed since v36 that could impact an app in production.

Tips

• Combine with websearch=true: the search finds the URLs, fetch reads them in depth.
• The result stays in the note: your research becomes a dated history.