Skip to main content

Advanced Web Crawling Platform with Deep Analysis and MCP Server

Project description

Crawilfy MCP Server

Crawilfy MCP Server — free, AI-native web crawling for the agent era

python 3.10+ · pypi 1.1.0 · MIT · MCP-native · code style black

A free, open-source MCP server that gives any agent (Claude Code, Cursor, Windsurf, …) 55 production-grade tools for the full web-crawling stack: deep analysis, stealth, API discovery, session recording → runnable crawler, smart extraction. No proprietary API. No per-request fee.

Crawilfy capabilities — deep analysis, stealth, record→crawler, smart extraction


Quick start

Three install paths — uvx, pipx, pip

The recommended path needs no install — uvx runs straight from PyPI:

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

Requires uv. Install once: curl -LsSf https://astral.sh/uv/install.sh | sh. Or use pipx run crawilfy-mcp-server / pip install crawilfy-mcp-server instead.

Where to put that JSON: Cursor → Settings → MCP. Claude Code → ~/.config/claude/mcp_settings.json. Windsurf → Settings → MCP Servers.


How it works

Agent → Crawilfy → Browser/HTTP/Proxy → target web

Your agent talks to Crawilfy over the Model Context Protocol. Crawilfy owns a hardened browser pool, an HTTP stack with TLS fingerprinting, and a rotating proxy pool. While it fetches pages, it captures network traffic, reads scripts, and introspects schemas — so the agent gets clean structured data, not raw HTML.


What's in the box

A short list — see the source for the full set of 55 tools.

Group Tools
Deep analysis deep_analyze, discover_apis, introspect_graphql, analyze_websocket, analyze_auth, detect_protection, detect_technology
Stealth stealth_request, configure_proxies, configure_rate_limit, add_proxy, test_proxy
Record → crawler record_session, stop_recording, export_recording, generate_crawler
Extraction smart_extract, extract_article, extract_tables, extract_links, extract_forms, extract_metadata, convert_to_markdown
Page interaction take_screenshot, fill_form, wait_and_extract, compare_pages, measure_performance, check_accessibility, get_dom_tree
Sessions & cache save_session, load_session, get_cookies, get_storage, clear_cache, get_cache_stats
Advanced (opt-in) execute_js, execute_cdp, deobfuscate_js, extract_from_js, solve_captcha

Smart extraction — bring any LLM, including free ones

smart_extract works without any API key using pattern matching. Plug in any OpenAI-compatible endpoint for higher accuracy — including FREE tiers:

# OpenRouter (free models exist)
CRAWILFY_LLM_PROVIDER=openrouter
CRAWILFY_LLM_API_KEY=sk-or-v1-xxx
CRAWILFY_LLM_MODEL=meta-llama/llama-3.2-3b-instruct:free

# Groq (free, very fast)
CRAWILFY_LLM_PROVIDER=groq
CRAWILFY_LLM_API_KEY=gsk_xxx

# Local Ollama (no key needed)
CRAWILFY_LLM_PROVIDER=ollama
CRAWILFY_LLM_MODEL=llama3.2

Together, DeepSeek, Mistral, Fireworks, and standard OpenAI also work via CRAWILFY_LLM_BASE_URL.


Configuration

Variable Default Notes
CRAWILFY_HEADLESS true Run browser without UI
CRAWILFY_BROWSER chromium chromium / firefox / webkit
CRAWILFY_POOL_SIZE 5 Max concurrent browsers
CRAWILFY_NAV_TIMEOUT 30.0 Page-load timeout (s)
CRAWILFY_API_KEY unset If set, every tool call must include matching _api_key
CRAWILFY_ALLOW_DANGEROUS_JS false Required for execute_js / execute_cdp / deobfuscate_js
CRAWILFY_JS_MAX_LENGTH 50000 Length cap for JS payloads
CRAWILFY_JS_EXEC_TIMEOUT 10.0 Per-script timeout (s)

Security

execute_js, execute_cdp, and deobfuscate_js are disabled by default — they execute or operate on arbitrary code in a real browser. Enable on trusted networks with CRAWILFY_ALLOW_DANGEROUS_JS=true. Even then, payloads are length-capped, time-bounded, and a denylist rejects eval, new Function, dynamic import(), document.write, importScripts, and WebAssembly.{compile,instantiate}. Set CRAWILFY_API_KEY so MCP clients must present a matching _api_key.

These are mitigations, not a sandbox: do not expose this server to untrusted clients.


Develop

git clone https://github.com/razavioo/crawilfy-mcp-server.git
cd crawilfy-mcp-server
make dev-install      # editable install + dev/captcha/ocr extras + pre-commit
make test             # pytest
make lint             # ruff + mypy

PRs welcome. Particularly interested in: distributed mode (Redis queue), result sinks (Postgres / S3), Prometheus metrics. See MIT License.

Made by emad.dev

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

crawilfy_mcp_server-1.1.2.tar.gz (143.5 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

crawilfy_mcp_server-1.1.2-py3-none-any.whl (122.0 kB view details)

Uploaded Python 3

File details

Details for the file crawilfy_mcp_server-1.1.2.tar.gz.

File metadata

  • Download URL: crawilfy_mcp_server-1.1.2.tar.gz
  • Upload date:
  • Size: 143.5 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.12.3

File hashes

Hashes for crawilfy_mcp_server-1.1.2.tar.gz
Algorithm Hash digest
SHA256 9d57f30d21136d9552d268d2e4637c9ac5fe000df2ffc6f192029fabf0cbee84
MD5 3728ac090de4a91091b6c2f414aaf384
BLAKE2b-256 7aa24cc7d82eb2b601095f64d7c538806bcee525728a7e43d45a8a94a8294bd0

See more details on using hashes here.

File details

Details for the file crawilfy_mcp_server-1.1.2-py3-none-any.whl.

File metadata

File hashes

Hashes for crawilfy_mcp_server-1.1.2-py3-none-any.whl
Algorithm Hash digest
SHA256 bf5d9b0aaacab4b658dd41ec31f8fec7a326c777ffc7a842cbf337084c9710ab
MD5 c627816838f34dd62c33a5b0c743e787
BLAKE2b-256 1308a2005953584dbfe64b35b54fe8aa988d5ecb1db2be945e9f09af0ca5d219

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page