MCP · fetch-url

fetch-url

Fetch and read any webpage from a natural language prompt.

MCP endpoint https://next.hal9.ai/mcp/mcpbuild/fetch-url

First time connecting an MCP? See Installing MCPs for setup steps.

What it does

fetch-url reads any webpage and returns its content as clean plain text. Scripts, navigation, and boilerplate are stripped automatically, leaving just the content your agent needs. No API key required — connect the endpoint and it works immediately:

fetch https://modelcontextprotocol.io

read the content at https://docs.anthropic.com/mcp

How agents use this MCP

Receive a URLe.g. "read https://example.com/docs/api-reference"
Fetch the pageHTTP GET with a standard User-Agent, follows redirects
Strip and returnRemoves scripts, styles, nav, and footer; returns clean paragraph text

Typical uses: letting an agent read documentation, check a page before recommending it, or extract information from a public web page without leaving the conversation.

Self-host or run locally

cd fetch-url
pip install -r requirements.txt
echo "fetch https://modelcontextprotocol.io" | python app.py

Variable	Required
FETCH_MAX_CHARS	No — defaults to 4000
FETCH_TIMEOUT	No — defaults to 15 seconds

Full source: github.com/hal9ai/mcp.build/tree/main/fetch-url

Connecting from Claude, Claude Code, Cursor, or another MCP client? Point it at the endpoint above — full setup instructions on the Installing MCPs page.