fastCRW (crw-mcp)

Name: fastCRW (crw-mcp)
Author: fastCRW

Web scraping, crawling, URL mapping and search for agents, run locally or through the fastcrw.com API

Official · fastCRW Rust Remote (hosted) · Streamable HTTP Local · stdio, Streamable HTTP Browser automation Web search

GitHub npm Website

55 popularity score ★ 1.2k ↓ 881/wk

Connect your client

Remote (hosted)

Add in one click

Add to Cursor Add to VS Code Add to LM Studio Add to Goose

Opens the app, which asks before it adds the server.

Or add it by command or config

Its docs show no set-up for a particular client: add this address in your client the way it adds a remote MCP server.

Addressits hosted endpoint

https://fastcrw.com/mcp

Local (once it is running)

Add in one click

Add to Cursor Add to VS Code Add to LM Studio Add to Goose

Opens the app, which asks before it adds the server.

Or add it by command or config

Claude CodeTerminal

claude mcp add --transport http fastcrw-crw http://localhost:3000/mcp

Cursor~/.cursor/mcp.json

{
  "mcpServers": {
    "fastcrw-crw": {
      "url": "http://localhost:3000/mcp"
    }
  }
}

VS Code.vscode/mcp.json

{
  "servers": {
    "fastcrw-crw": {
      "type": "http",
      "url": "http://localhost:3000/mcp"
    }
  }
}

Codex~/.codex/config.toml

[mcp_servers.fastcrw-crw]
url = "http://localhost:3000/mcp"

Gemini CLITerminal

gemini mcp add --scope user --transport http fastcrw-crw http://localhost:3000/mcp

Local (npm)

Add in one click

Add to Cursor Add to VS Code Add to LM Studio Add to Goose

Opens the app, which asks before it adds the server.

Or add it by command or config

Claude CodeTerminal

claude mcp add fastcrw-crw -- npx -y crw-mcp

Claude Desktopclaude_desktop_config.json

{
  "mcpServers": {
    "fastcrw-crw": {
      "command": "npx",
      "args": [
        "-y",
        "crw-mcp"
      ]
    }
  }
}

Cursor~/.cursor/mcp.json

{
  "mcpServers": {
    "fastcrw-crw": {
      "command": "npx",
      "args": [
        "-y",
        "crw-mcp"
      ]
    }
  }
}

VS Code.vscode/mcp.json

{
  "servers": {
    "fastcrw-crw": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "crw-mcp"
      ]
    }
  }
}

Codex~/.codex/config.toml

[mcp_servers.fastcrw-crw]
command = "npx"
args = ["-y", "crw-mcp"]

Gemini CLITerminal

gemini mcp add --scope user fastcrw-crw npx -- -y crw-mcp

Settings 8

`CRW_API_URL`	Optional: Remote CRW server to forward calls to (https://api.fastcrw.com for the cloud); switches to proxy mode.
`CRW_API_KEY`	Optional: Bearer key for the fastcrw.com cloud or a secured self-hosted server.
`CRW_LOCAL`	Optional: Truthy value forces embedded mode and ignores any configured API URL.
`CRW_RENDERER__LIGHTPANDA__WS_URL`	Optional: WebSocket address of a CDP browser for JavaScript rendering in embedded mode.
`CRW_CONFIG`	Optional: Path to a config file (embedded mode only).
`CRW_MCP__HIDE_CREDITS`	Optional: Strips credit-cost fields from tool results.
`CRW_MCP_BINARY`	Optional: Path to a crw-mcp binary the npm launcher runs instead of downloading one.
`RUST_LOG`	Optional: Log level; logs go to stderr.

What it does

crw-mcp is the MCP server for CRW, a web scraping engine written in Rust. It turns a page into markdown, HTML, a link list or schema-shaped JSON, follows a bounded crawl across a site, lists a site's URLs from its sitemap and a short link walk, runs web searches, and converts PDFs to markdown.

It runs in one of two modes. In embedded mode, the default, the scraping engine runs inside the MCP process on your machine and needs no account. In proxy mode, chosen when an API URL is found (the --api-url flag, CRW_API_URL, CRW_CLIENT__API_URL, or client.api_url in ~/.config/crw/config.toml written by `crw setup`), every tool call is forwarded to a remote CRW server: the paid fastcrw.com cloud or your own self-hosted crw-server. Setting CRW_LOCAL to a truthy value pins embedded mode even when a config file names an API URL. A running crw-server also serves MCP over HTTP at /mcp, and fastcrw.com runs a hosted endpoint at https://fastcrw.com/mcp that signs in with OAuth (scope mcp:tools, per its protected-resource metadata).

The npm package is a small Node launcher. It looks for a binary named in CRW_MCP_BINARY, then a prebuilt platform package, and otherwise downloads the matching release archive from GitHub on first run, refuses it unless it matches the release's SHA256SUMS, and caches it under ~/.cache/crw-mcp/v<version> (%LOCALAPPDATA% on Windows) (bin/crw-mcp.js at v0.37.2). Its `install` subcommand instead registers the server and a CRW skill file in the AI tools it detects.

What it needs

To run it yourself	Node.js

Node.js for npx; the launcher runs a prebuilt binary for macOS, Linux or Windows on x64 or arm64. Embedded mode needs no account. JavaScript rendering locally needs a CDP browser such as LightPanda (CRW_RENDERER__LIGHTPANDA__WS_URL); without one it fetches over plain HTTP. Cloud mode needs a fastcrw.com account and API key (CRW_API_URL plus CRW_API_KEY); calls spend credits. crw_extract and json/summary output need an LLM, configured on the server or passed as a key.

Limits

In embedded mode crw_search is listed only when a search backend is configured, so most local setups have 8 tools; proxy mode always lists all 9. Content fields are cut to about 15,000 characters by default (maxLength: 0 lifts it) and crw_map returns up to 100 URLs by default (limit: 0 lifts it). crw_crawl stops at depth 2 and 10 pages by default. crw_parse_file has no OCR: scanned PDFs come back empty with a warning. Crawl and map follow robots.txt by default. Clicking and form filling are not in this server; the makers point to their separate crw-browse server for that.

Who it suits

Suits developers who want scraping and crawling tools without an account: embedded mode keeps fetching on their own machine and is free. Pick cloud mode for managed proxies, JavaScript rendering and always-on search, at a per-credit cost. Teams that already host crw-server can point the client at its /mcp endpoint. For multi-step browsing (logging in, clicking through flows) the makers' own crw-browse server fits better. AGPL-3.0 licensing matters to anyone embedding the engine in a product.

Tools 9

crw_scrape: Fetches one URL and returns markdown, HTML, links, a screenshot or schema-shaped JSON
crw_crawl: Starts a bounded crawl from a URL and returns a job id
crw_check_crawl_status: Polls a crawl job and returns the pages collected so far
crw_map: Lists a site's URLs from its sitemap and a short link walk without scraping them
crw_extract: Starts a job that pulls structured JSON from URLs using a prompt or schema and an LLM
crw_check_extract_status: Polls an extract job for its status and results
crw_cancel_extract: Asks a running extract job to stop and returns its status
crw_parse_file: Converts a base64-encoded PDF to markdown, without OCR
crw_search: Searches the web and can scrape the top results; up to 20 results, 5 by default

Popularity

55 popularity score, on our list (the bar is 36)

★ 1.2k ↓ 881/wk

GitHub stars, whole repository	1,233	23.6 points
npm downloads, last week	881	23.6 points
Official server	Yes	8 points

It lives in the folder mcp/crw-mcp of fastcrw/crw, which holds more than this server. Its stars are the whole repository's, so in the score they never count for more than its own downloads.

Measured 2026-11-07. How the score works.

Where it is published

Code	fastcrw/crw/mcp/crw-mcp ★ 1.2k, the whole repository
npm	crw-mcp 0.42.1, 2026-11-06
Hosted	`https://fastcrw.com/mcp` Streamable HTTP · OAuth sign-in

What it can reach

Runs	On your machine, or on the maker's servers
Reaches	The internet, Drives a browser
Changes things	No: it reads only
Sign-in	OAuth sign-in for the hosted endpoint; none to run it yourself
Licence	AGPL-3.0 (npm)
Last code change	2026-11-06

Every tool except crw_parse_file fetches arbitrary URLs from the public web, and in embedded mode it does so from your machine and IP. crw_parse_file works on bytes passed in the call, not on file paths. No tool changes outside data, though crawl and extract create jobs, and in cloud mode each call spends account credits. In proxy mode the URLs and results pass through fastcrw.com or whichever server CRW_API_URL names, and a URL saved by `crw setup` in ~/.config/crw/config.toml turns proxy mode on unless CRW_LOCAL is set. On first run without a platform package the launcher downloads a binary from GitHub releases and checks it against SHA256SUMS.

Over time

GitHub stars, whole repository 1.2k +102 (+9 %)

1.2k1.1k

15 Oct 20267 Nov 2026

npm downloads a week 881 +639 (+264 %)

1.3k35

6 Apr 20267 Nov 2026

Before the site measured it, read from npm's daily download counts.

Popularity score 55 +14 (+35 %)

5540

15 Oct 20267 Nov 2026

What is rising across the directory

Sources

Page checked 2026-10-15.

Badge for your README

Shows this server's rank in its category, and stars for a server that runs on your machine. It updates every hour.

MarkdownREADME.md

[![fastCRW (crw-mcp) on EveryMCPServer](https://everymcpserver.com/badge/crw-web-scraper-2.svg)](https://everymcpserver.com/servers/crw-web-scraper-2)

Connect your client

Add in one click

Or add it by command or config

Add in one click

Or add it by command or config

Add in one click

Or add it by command or config

Settings 8

What it does

What it needs

Limits

Who it suits

Tools 9

Similar servers

Hyperbrowser MCP Server

Firecrawl MCP Server

Bright Data MCP

Crawl4AI

Scrapling MCP Server

SearXNG MCP Server