Repository intelligence

feder-cr/Jobs_Applier_AI_Agent_AIHawk

An open-source MCP server that drives a stealth, patched Firefox browser, allowing AI clients to interact with web pages, fill out forms, and navigate multi-step workflows without triggering anti-bot challenges.

CLOUDM0N decision

MUST TRY

Trust PASS · 100/100

Good fit if

AI agents and developers using MCP-compliant clients (such as Claude Code, Claude Desktop, or Cursor) who need to automate complex web-browsing tasks.

Watch out for

Does not support macOS, as the engine is built specifically for Windows and Linux.

Practical intelligence

What matters before you adopt it

Decision first · detail on demand

Problem it solves

The high failure rates and frequent blocks encountered by standard headless browser automations when modern anti-bot systems detect fingerprint contradictions or runtime scripting footprints.

Best for

AI agents and developers using MCP-compliant clients (such as Claude Code, Claude Desktop, or Cursor) who need to automate complex web-browsing tasks.

Users needing to perform background web automation without getting blocked by Cloudflare, DataDome, or other anti-bot intermediaries.

Main trade-offs

Does not support macOS, as the engine is built specifically for Windows and Linux.

Does not contain a built-in automated captcha solver for interactive or visual challenges.

Provides no direct disk or filesystem access, meaning the AI client must handle file inputs/outputs via its own tools.

Why it stands out

Ensures complete coherence across proxy geolocation, timezone, clock, locale, and hardware strings to minimize anomaly risk scores.

Exposes a compact, 13-tool set tailored specifically to what a browser needs without overloading the LLM context window.

Features a browser snapshot tool that returns only visible interactive elements instead of the entire accessibility tree, preventing attention flood.

Trust & CVEs

Security evidence without the noise

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

Security findings

—

Scanner unavailable

Critical

—

High

—

Medium

—

Low

—

View trust evidence & security findings▼

Why this score

No trust rationale was stored for this scan.

CLOUDM0N findings

Architecture from code9 modules · 11 edges▼

Structural evidence

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

Code files

Modules

Dependency edges

Core modules

(root)

3 files

src

5 files

src/libs

library code

20 files

src/resume_schemas

2 files

src/utils

utility helpers

3 files

Dependency flow

(root)→src

(root)→src/libs

(root)→src/resume_schemas

(root)→src/utils

src→(root)

src/libs→(root)

src/libs→src

src/libs→src/utils

src/resume_schemas→src

src/utils→src

Detected languages

Python

Detected frameworks

LangChain · OpenAI SDK

Architecture evidence details

flowchart TD
    %% repo — high-level architecture (DRAFT, refine me)
    n0["(root) · 3 files"]
    n1["src · 5 files"]
    n2["src/libs · library code · 20 files"]
    n3["src/resume_schemas · 2 files"]
    n4["src/utils · utility helpers · 3 files"]
    n5["tests · tests · 1 file"]
    n0 --> n1
    n0 --> n2
    n0 --> n3
    n0 --> n4
    n1 --> n0
    n2 --> n0
    n2 --> n1
    n2 --> n4
    n3 --> n1
    n4 --> n1
    n5 --> n0
    class n2,n4 shared
    class n5 test
    classDef shared fill:#79706e,color:#ffffff,stroke:#5d5654
    classDef test fill:#499894,color:#ffffff,stroke:#397975

Evidence, security & integrations▼

Integrations

MCP (Model Context Protocol)Claude CodeClaude DesktopCursoruv / uvxFirefoxPlaywright (via invisible_playwright)lib_resume_builder_AIHawk

Security notes

• Does not allow the browser or server direct access to your local disk, requiring files to be handled by your client's native filesystem tools.

• Enables persistence of active logins and cookies across restarts when configured with a local STEALTHFOX_PROFILE_DIR folder.

• Derived profile values (such as screen metrics and GPU strings) can be locked to a steady state via STEALTHFOX_SEED to avoid signaling identity shifts.

Still unknown

• The README does not specify if the source-level patched Firefox binary can be run on macOS or if macOS support is planned.

• No performance benchmarks regarding CPU, RAM, or maximum concurrent tab limits are provided for the patched Firefox engine.

• The complete list of all thirteen individual tools exposed by the MCP server is not exhaustively listed or defined.

Adoption guidance▼

Adopt if

+ You use an MCP-enabled AI client on Windows or Linux and need to automate web workflows such as job applications or multi-step travel searches.

+ You want to reduce anti-bot interruptions and Cloudflare/DataDome challenges through a source-level patched Firefox browser.

Avoid if

– You are running on macOS or other non-Windows/Linux host environments.

– You require a fully automated, standalone job applier command-line tool, as the original 2024 version is unmaintained.

– You need a tool that can autonomously solve interactive visual captchas (such as selecting motorbikes).

How it works & getting started▼

How it works

1.The user configures their preferred MCP client to register the invisible-playwright-mcp server using uvx.

2.The server resolves the exit IP of the configured proxy before launching the browser.

3.The system derives coherent timezone, geolocation, locale, and language properties from the proxy IP.

4.It seeds a source-level patched Firefox browser to keep hardware metrics like GPU and screen strings steady.

5.The AI client invokes the 13 exposed tools to parse layouts, fill forms, click buttons, or extract tables.

Getting started

→ Ensure your environment has Python 3.11+ and uv installed.

→ Add the invisible-playwright-mcp runner configuration to your MCP client's settings file.

→ To use a proxy, include the --env flag pointing to STEALTHFOX_PROXY during execution.

→ Restart your MCP client and query a fingerprint analysis page to verify that public IP, locale, and timezone properties are perfectly aligned.

Agent handoff

Use with any agent

JSON API

Alternatives

Nearby repositories worth comparing before adoption.

Compare top options →

camofox-browser

redf0x1/camofox-browser

Fit

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

Trust REVIEW · 60

Compare →

veilbrowser

acunningham-ship-it/veilbrowser

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 →

nanobrowser

nanobrowser/nanobrowser

Fit

Open-Source Chrome extension for AI-powered web automation. Run multi-agent workflows using your own LLM API key. Alternative to OpenAI Operator.

Trust REVIEW · 90

Compare →