tilion 0.1.0

Open stealth-browser framework — drive a real recompiled-Chromium (Fortress) that defeats bot-detection fingerprint suites. Drop-in stealth for browser-use, Playwright, Crawl4AI, with an MCP server for AI agents.

Tilion

A stealth browser for AI agents. pip install tilion — drive a real, undetected Chromium that gets past Cloudflare, DataDome, and bot detection, with no server, no account, and no API key.

Framework & MCP: Beta · runs local & free · Tilion Cloud (residential egress) coming soon · 📖 Full documentation · 📊 Benchmarks

Real, dated run: stockx.com (PerimeterX). A stock browser gets HTTP 403 — "Access denied"; Claude with the Tilion MCP returns clean JSON. Reproduce: python demo/make_split_demo.py.

Tilion is an open stealth-browser framework. It runs a recompiled-Chromium stealth engine (Fortress) locally, in-process, and gives you one clean API for fetching protected pages, extracting content, crawling sites, reconnaissance, and multi-step automation — plus a Model Context Protocol (MCP) server so AI agents can reach for it the moment they get blocked.

pip install tilion

Quickstart

import asyncio
from tilion import Tilion

async def main():
    async with Tilion() as t:                     # boots the Fortress stealth engine locally
        page = await t.fetch("https://protected.example")   # past Cloudflare/DataDome/403
        print(page["title"], page["text"][:200])

        data = await t.extract("https://site/pricing")       # clean markdown + tables
        docs = await t.crawl("https://site", depth=2)        # whole-site crawl (auto-SPA)
        apis = await t.recon("https://site")                 # discover the site's private API
        hits = await t.search("undetected playwright")       # real-browser web search

asyncio.run(main())

No uvicorn, no Redis, no auth — local mode holds one real browser and drives it directly.

Drop-in stealth for your existing stack

Already using browser-use, Playwright, Puppeteer, Crawl4AI, or Stagehand? They all connect to a browser by CDP URL. Point that at Tilion's Fortress engine and your existing code runs through the stealth browser — one line, no rewrite:

from tilion import Tilion

t = await Tilion().start()
cdp_url = t.cdp_url            # hand this to any CDP-speaking stack

# browser-use:
from browser_use import Agent, BrowserSession
agent = Agent(task="...", browser_session=BrowserSession(cdp_url=cdp_url))

# Playwright:
browser = await playwright.chromium.connect_over_cdp(cdp_url)

MCP server — stealth browsing for any AI agent

Ship the Fortress MCP server so a coding/browsing agent auto-selects it when a fetch is blocked:

pip install "tilion[mcp]"
tilion-mcp          # or: python -m tilion.mcp

Claude Desktop / Cursor — add to your MCP config:

{ "mcpServers": { "fortress": { "command": "tilion-mcp" } } }

26 intent-loaded tools including: fetch_protected_page, extract_page, crawl_site, recon_site_apis, search_web, run_browser_task, screenshot_page, save_page, list_browser_tasks, get_stealth_cdp_endpoint. Pre-warmed on startup (~100 ms first call), concurrency-safe, timeout- and SSRF-guarded.

What you get

fetch	stealth GET past Cloudflare/DataDome/403 + auto challenge-resolve
extract	page → clean markdown + tables + metadata (or a schema-shaped record)
crawl / spa_crawl	whole-site crawl, auto-handles SPA/JS + lazy-load → sitemap
recon	reverse-engineer a site's private XHR/JSON API (secret-scrubbed)
search	real-browser web search (no SERP API)
agent	20 multi-step flows: login, paginate, infinite-scroll, checkout, downloads…
screenshot / save	PNG, or export as PDF / HTML / text
cdp_url	raw CDP endpoint for browser-use / Playwright / Puppeteer
t.page(url)	escape hatch to the full DOM API (click, type, wait, evaluate, …)

How stealth works

The engine is a recompiled Chromium C++ fork, not a JS patch or a stealth plugin. Persona, User-Agent, WebGL, and canvas fingerprints are applied natively with genuine binding returns (toString() === [native code]), so there are no JavaScript injection tells. In independent fingerprint suites it runs Sannysoft-clean, CreepJS 0% headless, browserscan "no bots."

Honest note: a great fingerprint on a datacenter IP still gets blocked by the biggest sites — real anti-bot decisions weigh the egress IP heavily. For production scraping of hostile targets, add a residential proxy (Tilion's proxy layer is pluggable) or use hosted Tilion cloud egress. Tilion does not claim to be "undetectable."

Install options

pip install tilion              # core: stealth browser + fetch/extract/crawl/recon/search
pip install "tilion[mcp]"       # + the MCP server
pip install "tilion[vision]"    # + LLM-driven agent (Claude/OpenAI/Gemini)
pip install "tilion[server]"    # + the multi-tenant control-plane API (self-host cloud)

License

Open-core. The funnel — tilion.facade + tilion.mcp — is BSD-3-Clause. The engine (tilion.core driver + the Fortress C++ binary via tilion-fortress) is proprietary, shipped as bytecode/binary. See OPENSOURCE_SPLIT.md.