Skip to content

MCP setupWorkflow

Fetch MCP Server

Let your agent download a web page and read it as clean markdown, in chunks, using the reference fetch server from the MCP project.

npx workspacetemplates add fetch-mcp --tool claude-code

Writes 1 file to your project root

.mcp.jsonmerge into your file10 lines · 122 BFile
{
  "mcpServers": {
    "fetch": {
      "command": "uvx",
      "args": [
        "mcp-server-fetch"
      ]
    }
  }
}

What it does

The fetch server adds one tool, fetch, that takes a URL and returns the page as markdown. You can ask for a longer slice with max_length and start_index, or request the raw HTML.

It is a Python package started with uvx, so you need uv installed. There is no account and no key.

By default it refuses private and loopback addresses and obeys a site's robots.txt for requests the model makes, which keeps an agent from poking at your internal network.

Example

you › Fetch https://example.com/docs/migration and give me the steps in order.

agent › The page came back as markdown in one chunk. Steps: 1. Update the config key. 2. Rename the two deprecated methods. 3. Run the codemod. 4. Rebuild. The page has a note that step 3 is optional.

Questions

Why did a fetch fail with a robots.txt error?
The server checks robots.txt for model-initiated fetches and refuses when the site disallows it or the file cannot be reached. The flag --ignore-robots-txt turns the check off, so use it only for sites you own.
Can I limit which sites it may visit?
Yes. Add --allowed-hosts followed by hostnames, such as docs.python.org or a wildcard like *.wikipedia.org, to the args.
How do I read a very long page?
The default cap is 5000 characters per call. Ask the agent to call fetch again with a higher start_index to continue where it stopped.
  • Asana MCP Server

    MCP setupWorkflow

    Free
  • Atlassian MCP Server

    MCP setupWorkflow

    Free
  • Brave Search MCP Server

    MCP setupWorkflow

    Free
  • ClickUp MCP Server

    MCP setupWorkflow

    Free