Firecrawl MCP alternative: free, no API key, runs on your machine

• You need every page of a site: a crawl, or a map of its URLs.
• The pages are web apps that show nothing until their scripts run, or you must click and type before the content appears.
• You want structured data that matches a schema, extracted by a model on the server.
• You run large batches and want the fetches to come from the addresses of a provider, not from your own.

Anyfetch does none of these. It reads one page or several pages that you name, and it searches.

• Your agent searches and reads pages during its work, as Claude Code, Codex, or Cursor does. That is hundreds of small calls, and each one costs a credit in a hosted service.
• You want no API key to create, store, and rotate, and no monthly limit that stops the agent in the middle of a task.
• You do not want the URLs that your agent reads to pass through another company.
• You want speed. Anyfetch reads one page in about 0.19 seconds with 17 MB of memory, because no browser starts.
• You want full Reuters and Bloomberg articles, Reddit threads with every comment, and Stack Overflow answers, from the data that these sites send.

{
  "mcpServers": {
    "firecrawl": {
      "command": "npx",
      "args": ["-y", "firecrawl-mcp"],
      "env": { "FIRECRAWL_API_KEY": "fc-..." }
    }
  }
}

{
  "mcpServers": {
    "anyfetch": { "command": "npx", "args": ["-y", "@aliasmode/anyfetch", "mcp"] }
  }
}

You can keep both servers. Let the agent use Anyfetch for search and for single pages, and Firecrawl for crawls. To install Anyfetch as a program and connect every agent on the machine, use the installer:

curl -fsSL https://aliasmode.github.io/anyfetch/install.sh | sh

irm https://aliasmode.github.io/anyfetch/install.ps1 | iex