Repository avatar
AI Tools
v3.6.0
active

webskrap

io.github.kacigaya/webskrap

Stealth web fetch, search and real Chrome control for AI agents.

Documentation

WebSkrap logo

WebSkrap

Async-first Python scraping on its own Chrome DevTools Protocol client, with persistent sessions, resource routing, stealth defaults, web search, a CLI, and an MCP server for agents.

WebSkrap version on PyPI Python 3.11 or newer Apache 2.0 License

Play the 24-second WebSkrap demo video: real Chrome passes public bot tests, and an AI agent drives it through the WebSkrap MCP server.

Install

pip install webskrap
webskrap install

webskrap install downloads Chrome for Testing. WebSkrap drives it through its own CDP client, with no Playwright. Upgrading from 2.x? Read the migration guide.

Quickstart

import asyncio

from webskrap import WebSkrapClient


async def main() -> None:
    async with WebSkrapClient() as client:
        result = await client.fetch("https://example.com")
        print(result.status, result.title)
        print(result.text[:200])


asyncio.run(main())

For deferred content, pass ready_selector="#content" to a Python fetch or --ready-selector '#content' to CLI fetch. Both MCP fetch tools accept ready_selector too. The selector must become visible after consent dismissal. Known challenge pages (DataDome, Cloudflare, Akamai, AWS WAF) and provider block pages (DataDome, Cloudflare, Akamai) raise blocked. WebSkrap waits for device checks that reload the page. It does not retry refusals or solve CAPTCHAs. Ordinary HTTP errors still return ok=False. For an empty HTTP error body, Chromium can show its own error page. The returned text or HTML is that rendered page. The status, headers, and final_url still come from the HTTP response.

For manual CAPTCHA completion, open a persistent browser with webskrap browser open --headed --session manual --proxy http://localhost:8888 URL. Complete the check in that browser. Keep the same session, profile, and proxy for later commands. On a remote host, use a desktop connection to its display. virtual_display=True alone does not provide a remote desktop.

Keep profile_mode="native" for the browser's own identity. On Linux with Xvfb, virtual_display=True runs a headed browser without a visible desktop window. Full profile emulation is useful for device tests, but cannot replace every worker or iframe value. See the identity limits.

Python batches

Use client.fetch_many(urls) to fetch a finite batch in order with one browser:

async with WebSkrapClient() as client:
    results = await client.fetch_many(
        ["https://example.com", "https://example.org"], text_only=True
    )

Each batch shares cookies and storage, then closes its browser. Separate batches and client.fetch() calls use fresh sessions. An explicit user_data_dir can still share state on disk. Empty batches do not launch a browser. The first page error stops the batch and raises; no partial list is returned. Ordinary HTTP errors remain results with ok=False. screenshot=True uses a separate generated PNG filename for each page.

For large inputs, per-page options, or state across batches, create a session with session = await client.session("crawl"). Call await session.fetch(url) for each URL. Close it with await client.close_session("crawl"), or let the client context close it. Session names belong to one client. Use separate sessions for unrelated callers.

Select elements

Query fetched HTML with CSS selectors. Use parse_html(html) for HTML from other sources.

async with WebSkrapClient() as client:
    result = await client.fetch("https://example.com")
    print(result.css_first("h1").text)
    print([a.attrs["href"] for a in result.css("a[href]")])

Search

A search loads the engine's results page in the same stealth browser and returns the organic hits. Click-tracking redirects are unwrapped, so each url is the destination itself. Bing is the default; use engine="ddg" for DuckDuckGo's HTML endpoint. Google is not offered.

async with WebSkrapClient() as client:
    result = await client.search("example domain", max_results=5)
    for hit in result.hits:
        print(hit.title, hit.url)
webskrap search "example domain" --engine ddg --max-results 10 --format json

When an engine distrusts the exit address it serves a bot challenge. That comes back as a blocked error, not as zero hits. Switch engine, reuse a persistent profile, or change the exit IP. WebSkrap does not retry and does not solve CAPTCHAs.

Visual browser interaction

Agents can see the rendered viewport and interact by coordinates, including canvas and controls missing from the accessibility tree. MCP browser_view returns a PNG image directly; browser_mouse clicks, moves, scrolls, or drags. browser_insert_text inserts literal text into the focused control.

webskrap browser open https://example.com
webskrap browser view viewport.png --format json
# Open the returned PNG with your agent's image viewer, then choose coordinates.
webskrap browser mouse click 120 80
webskrap browser insert-text "hello"
webskrap browser press Enter
webskrap browser view viewport.png --format json
webskrap browser close

Coordinates are CSS pixels from the original image's top-left. Take a fresh view after scrolling or page changes. Headless Chromium needs no Xvfb for this; the agent's client and model must support image input.

Load only what the task needs

The default resource policy is auto. Text-only fetches skip images, fonts and media. HTML fetches, screenshots and interactive pages load all resources. Set resource_policy="all", "lite", or "documents" to override this choice. Use documents only when removing CSS preserves the required result. Fetches use domcontentloaded; set ready_selector for deferred content. Use browser_find(text="Save") to locate a known control with ancestor context. Use browser_snapshot(target="e15", depth=2) for one subtree; add boxes=true only when geometry matters. Reuse returned refs with browser_interact. Use networkidle, screenshots and full snapshots only when the task needs them. CLI and MCP reject cookies by default. Disable this only for known pages where it is unnecessary. In Python, enable rejection with SessionConfig(decline_cookies=True).

Documentation

The agent skill is available at .agents/skills/webskrap/SKILL.md, .claude/skills/webskrap/SKILL.md, and .codex/skills/webskrap/SKILL.md. The file lives in .agents/skills/webskrap/. The .claude and .codex folders link to it, and every folder links to the root references/. The checkout must support symbolic links. On Windows, enable Developer Mode and Git's core.symlinks before cloning.

Security

Persistent browser sessions keep Chromium's OS sandbox, store their profiles 0700, and confine MCP screenshot output to ./webskrap-output. See SECURITY.md for the security model and how to report a vulnerability, and CHANGELOG.md for what changed.

Contributing

See CONTRIBUTING.md for setup, checks, and pull request expectations.

PYPI
webskrap
Install Command
pip install webskrap
Runtime: uvx