PhantomPool

Browser pool built around a simple idea:

if you’re going to run real browsers, you need control — not just concurrency.

I built this because most Cloudflare bypass tools I found either:

• don’t scale,
• spawn browsers per request,
• or completely fall apart under concurrent load.

PhantomPool focuses on orchestrating browsers, not abusing them.

---

What it does

• Keeps a persistent pool of real Chromium instances
• Uses leases to guarantee exclusive browser access
• Applies hard cancellation on timeouts and disconnects
• Uses a bounded queue to apply backpressure

That’s it. No magic.

---

What it does not do

• No fingerprint spoofing
• No TLS / JA3 tricks
• No exploit-based bypasses

If the browser can’t load the page, PhantomPool won’t pretend it can.

---

How it works (high level)

HTTP request
  → Queue (bounded)
    → Worker
      → Browser lease
        → Chromium

Each request becomes a job. Each job needs a lease. Each lease owns a browser until it finishes or is cancelled.

---

Why leases instead of “just a pool”

Because without leases:

• browsers get reused while dirty
• cancellation becomes best-effort
• bugs turn into ghost states

Leases give PhantomPool:

• authority
• cleanup guarantees
• predictable failure modes

---

API

GET /navigate

Query:

• url (required)
• timeoutMS (optional)

Responses:

• 200 – success
• 429 – queue full
• 499 – client aborted
• 5xx – internal error

---

Configuration

Everything is configured via env vars.

HOST=0.0.0.0
PORT=4000

BROWSER_POOL_SIZE=1
BROWSER_POOL_RETRYMS=1000
BROWSER_LEASE_TIMEOUTMS=30000

QUEUE_MAXJOBS=20
QUEUE_RETRYMS=10

Defaults are conservative on purpose.

---

Docker

docker build -t phantompool .
docker run -p 4000:4000 phantompool

Chromium is bundled in the image.

---

Current state

This project is:

• stable under load
• intentionally minimal
• still evolving

Things that can improve:

• smarter scheduling
• metrics
• multiple browser types
• distributed workers

PRs are welcome.

---

License

MIT