CLOUDM0N
Repository intelligence

psyb0t/docker-stealthy-auto-browse

GitHub

A stealth browser automation platform that runs a custom Firefox build in Docker with zero Chrome DevTools Protocol exposure, real OS-level PyAutoGUI inputs, and control via a JSON HTTP API or MCP server.

CLOUDM0N decision
REVIEW BEFORE ADOPTION
Trust REVIEW · 80/100
Good fit if

AI agents and automated workflows that need to browse websites protected by modern anti-bot detectors.

Watch out for

OS-level system clicks can land in the wrong coordinates if the window offset becomes stale, requiring manual calibration.

Practical intelligence

What matters before you adopt it

Problem it solves

The high failure and blocking rates of standard automated browsers due to detectable Chrome DevTools Protocol (CDP) signals and synthetic, injected DOM actions.

Best for
AI agents and automated workflows that need to browse websites protected by modern anti-bot detectors.
Developers requiring robust browser automation with real mouse and keyboard event simulation instead of injected DOM actions.
Teams scaling web scraping pipelines across a clustered browser fleet with synchronized session cookies.
Main trade-offs
OS-level system clicks can land in the wrong coordinates if the window offset becomes stale, requiring manual calibration.
When running in cluster mode with multiple replicas, the HTTP API and MCP server restrict commands exclusively to script execution.
Virtual media dynamic updates are limited to local folder uploads and cannot load remote URLs or live streams.
Why it stands out
Successfully passes major bot detectors including Cloudflare, CreepJS, Pixelscan, and BrowserScan.
Simulates physical human interaction via PyAutoGUI mouse movements and keyboard events.
Clustering capabilities with instant fleet-wide cookie synchronization over Redis PubSub.
Trust & CVEs

Security evidence without the noise

Trust remains a decision signal; CVEs and scanner evidence explain what is driving the risk.

Security findings
2
CLOUDM0N scanner findings
Critical
0
High
0
Medium
2
Low
0
View trust evidence & security findings▼
Why this score
No trust rationale was stored for this scan.
CLOUDM0N findings
MEDIUM
The project can spawn operating-system processes; review command construction and input handling. 5 sample match(es) found.
MEDIUM
Dockerfile does not end with an explicit non-root USER.
Architecture from code7 modules · 0 edges▼
Structural evidence

Modules and dependency edges extracted from repository code. This is code evidence, not README inference.

Code files
19
Modules
7
Dependency edges
0
Core modules
app
11 files
scripts
1 files
Dependency flow
No dependency edges were extracted.
Detected languages
Shell · Python · JavaScript
Detected frameworks
—
Architecture evidence details
flowchart TD
    %% repo — high-level architecture (DRAFT, refine me)
    n0["app · 11 files"]
    n1["scripts · scripts · 1 file"]
    n2["tests · tests · 7 files"]
    class n1 infra
    class n2 test
    classDef infra fill:#b35c00,color:#ffffff,stroke:#8f4a00
    classDef test fill:#499894,color:#ffffff,stroke:#397975
Evidence, security & integrations▼
Integrations
CamoufoxXvfbPyAutoGUInoVNCOpenboxffmpegMCP (Model Context Protocol)Claude Code
Security notes
• Enforces access security using Bearer tokens via the AUTH_TOKEN environment variable.
• Dynamic media modes prevent path traversal by only accepting uploads within the configured media directory.
• Static security scanning can be triggered locally via make sec to produce a SARIF audit report.
Still unknown
• The README does not provide the exact single-instance docker run or pull command examples.
• It does not state the minimum system hardware resources (such as CPU cores or host RAM) required for performance.
• The specific structural database schema for the internal session tracking is not detailed.
Adoption guidance▼
Adopt if
+ You need to bypass modern, aggressive bot walls (like Cloudflare, CreepJS, or BrowserScan) during automated browsing operations.
+ You require native, direct integration between your AI coding agents and an MCP-compliant browser.
+ You need to automate browser flows that rely heavily on functional virtual camera or microphone inputs.
Avoid if
– You require automated, hands-free captcha-solving capabilities natively built into the platform.
– You need highly precise OS-level click targeting that remains perfectly reliable without potential coordinate offsets or calibration steps.
How it works & getting started▼
How it works
1.The application runs Camoufox, a custom Firefox fork without CDP exposure, inside a Docker container.
2.Xvfb establishes a virtual framebuffer to simulate a graphical monitor so the browser runs in a non-headless state.
3.PyAutoGUI generates real OS-level mouse movements and keyboard events to interact with pages.
4.Preloaded extensions automatically strip tracking parameters and handle cookie consent prompts.
5.Users or AI agents control the browser flow remotely by sending multi-step scripts to the HTTP API or the MCP endpoint.
Getting started
→ Connect to the HTTP API on port 8080.
→ Access the visual noVNC viewer at http://localhost:5900/.
→ Define automated workflows using YAML script configuration files or JSON payloads.
Agent handoff
Use with any agent
JSON API
Alternatives

Nearby repositories worth comparing before adoption.

Compare top options →
camofox-browser
redf0x1/camofox-browser
51
Fit

Anti-detection browser server for AI agents — REST API wrapping Camoufox engine with OpenClaw plugin support

Trust REVIEW · 60
Compare →
iphone-use
leeguooooo/iphone-use
41
Fit

Computer-use, but for the iPhone — AI agents (and your browser) see and drive a real phone over macOS iPhone Mirroring. Low-latency WebRTC video, near-native touch, HTTP API + MCP. Rust, self-hosted.

Trust REVIEW · 65
Compare →
veilbrowser
acunningham-ship-it/veilbrowser
42
Fit

Stealth browser for AI agents — real Chrome over raw CDP, no Playwright/Puppeteer. TypeScript + MCP-native. Passes sannysoft 57/57, bypasses Cloudflare.

Trust REVIEW · 70
Compare →