ApplyGenie MCP 🧞‍♂️

Your magical AI assistant that grants your job wishes.
A production-grade Model Context Protocol (MCP) server for desktop control, autonomous job application form filling, local application tracking, and personalized AI mock interview preparation.

---

🌟 What is ApplyGenie?

ApplyGenie bridges the gap between job discovery, form filling, application tracking, and interview readiness. It runs locally as an MCP server, empowering LLM agents (Claude Desktop, Cursor, Antigravity, VS Code, Windsurf) with 48 specialized tools:

1. 🖥️ Desktop Control & Vision: Fast screenshot captures (mss), DPI-corrected mouse navigation, human-like typing jitter (pynput), and native Windows window management (pywinauto).
2. 📋 Smart Candidate Profile: Local JSON schema storing your education, work experiences, skills, work authorizations, and custom portal answer banks.
3. 🌐 AI Browser Automation & Anti-Detection Form Filling: Playwright-powered stealth engine with deep CDP cloaking, native .toString() code spoofing, discrete WebGL GPU profiles (RTX 3080), deterministic canvas/audio fingerprint protection, Fitts's Law cubic Bézier biometric mouse trajectories, realistic keystroke dynamics with unicode currency fallback, and automated Challenge Radar (Cloudflare Turnstile, reCAPTCHA, hCaptcha). Parses forms across nested iframes & Shadow DOMs, maps candidate profiles, and requires explicit confirmation before submitting.
4. 📊 Application Tracker & Analytics: Local SQLite database logging every applied role, tracking interview conversion rates, and charting application velocity.
5. 🎤 Mock Interview Engine: Socratic coding hint ladders (4 rungs), behavioral STAR coaching ("I" vs "We" ownership ratio analysis), and 4-phase system design state machines.

---

🚀 Quickstart & Installation

Option 1: Run directly with uvx (Zero manual setup)

uvx applygenie-mcp

Option 2: Install via pip

pip install applygenie-mcp

# With full browser automation support:
pip install "applygenie-mcp[browser]"
playwright install chromium

---

⚙️ Configuration for MCP Clients

1. Claude Desktop

Add to your %APPDATA%\Claude\claude_desktop_config.json (Windows) or ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):

{
  "mcpServers": {
    "applygenie": {
      "command": "uvx",
      "args": ["applygenie-mcp"]
    }
  }
}

2. Google Antigravity

Add to your project's .agents/mcp_config.json or ~/.gemini/config/mcp_config.json:

{
  "mcpServers": {
    "applygenie": {
      "command": "python",
      "args": ["-m", "applygenie.server"],
      "cwd": "C:\\Personal Projects\\applygenie-mcp"
    }
  }
}

3. Cursor / Windsurf

Navigate to Settings > Features > MCP > Add New MCP Server:

• Name: ApplyGenie
• Type: command
• Command: uvx applygenie-mcp

---

🛠️ Complete MCP Tool Reference

1. Desktop Control & Vision

Tool	Description
screenshot(monitor=0)	Captures JPEG screenshot encoded in base64.
get_screen_size()	Enumerates all connected monitors and resolution bounds.
mouse_click(x, y, button, clicks)	Clicks mouse with DPI scaling correction.
mouse_move(x, y)	Moves mouse cursor smoothly to target coordinates.
mouse_scroll(direction, amount)	Scrolls mouse wheel up or down.
mouse_drag(start_x, start_y, end_x, end_y)	Performs click-and-drag between coordinates.
get_mouse_position()	Returns current mouse cursor coordinates.
type_text_input(text, interval)	Types text with randomized jitter (±30%) to emulate human cadence.
press_keyboard_key(keys)	Executes hotkeys (e.g. ctrl+a, tab, enter).
list_open_windows()	Returns all visible application windows.
focus_app_window(title)	Brings target window to foreground via fuzzy title match.
minimize_app_window(title)	Minimizes target window.
maximize_app_window(title)	Maximizes target window.
get_current_window()	Returns info about currently focused active window.
search_windows(pattern)	Finds windows matching regex or substring.
wait_seconds(seconds)	Pauses execution safely (capped at 30s).
get_action_history(last_n)	Inspects safety audit action logs.

2. Profile & Resume Management

Tool	Description
get_user_profile()	Loads complete candidate profile.
get_user_profile_summary()	Returns concise summary of skills, credentials, and work history.
setup_user_profile(...)	Initializes candidate personal info, URLs, location, and defaults.
update_profile_attribute(field, val)	Updates a specific top-level profile field.
add_work_experience(...)	Appends a work experience or internship entry.
add_education_entry(...)	Appends a degree / university qualification.
add_project_entry(...)	Appends a notable software project.
save_portal_custom_answer(pattern, ans)	Saves reusable answer for portal questions (e.g. "Why hire you?").
parse_resume_file(filepath)	Extracts text from PDF or DOCX resume.

3. Browser Automation & Form Filling

Tool	Description
browser_open(url, headless)	Launches stealth Chromium browser with deep CDP cloaking and discrete GPU profiles.
browser_navigate(url)	Navigates browser to destination page with automatic Challenge Radar auto-wait.
browser_get_page_text()	Extracts visible text content from current tab.
browser_get_page_screenshot()	Captures high-res screenshot of current webpage.
browser_get_accessibility_tree()	Extracts full ARIA tree snapshot to observe custom comboboxes, listboxes, and dialogs.
browser_detect_form_fields()	Parses inputs, textareas, selects, and file uploads across nested iframes & Shadow DOM.
browser_fill_detected_form(kinetic)	Auto-maps profile data and fills all detected inputs (with optional kinetic biometric typing & mouse movement).
browser_upload_resume(filepath)	Uploads resume PDF/DOCX to detected file input field.
browser_submit_form(confirm=True)	Clicks submit button (Requires explicit human confirmation).
browser_close()	Closes browser gracefully.

4. Application Tracker

Tool	Description
track_application(...)	Records a new application in local SQLite database.
update_application_progress(id, status)	Updates status (interview, screening, offer, rejected).
get_tracked_applications(...)	Lists applications with optional status/company filtering.
search_tracked_applications(query)	Full-text search across roles, companies, and notes.
get_application_analytics()	Computes interview conversion rates, offer rates, and top companies.

5. Mock Interview Engine

Tool	Description
start_mock_interview(type, company, diff)	Starts coding, behavioral, or system design session.
get_coding_hint()	Advances the 4-rung Socratic hint ladder without spoiling answers.
analyze_behavioral_star_answer(text)	Evaluates STAR structure and calculates "I" vs "We" ownership ratio.
advance_system_design_interview(input)	Steps through 4 architectural interview phases.
generate_interview_scorecard(...)	Produces structured 1-5 rubric report card and hiring recommendation.

---

🛡️ Safety & Failsafe Architecture

ApplyGenie enforces strict safety guardrails:

1. Mouse Corner Failsafe: Moving the cursor to screen corner (0, 0) immediately halts all automation.
2. Rate Limiting: Enforces a maximum action frequency (default: 30 actions/min) and minimum delays to prevent runaway loops.
3. Human-in-the-Loop Confirmation: Form submission and destructive actions require explicit confirm=True confirmation. The copilot fills 95% of the form, pausing for human review.
4. Local Data Storage: All candidate profile data, resume caches, and application records stay on your local disk at ~/.applygenie/. No telemetry, no cloud sync.
5. Key Combo Shield: Dangerous system combinations (alt+f4, ctrl+alt+del) are blocked by default.

---

🧪 Running Tests

python -m pytest tests/ -v

---

📄 License

This project is licensed under the MIT License — see the LICENSE file for details.

Developed with ❤️ by Tanishq Mangal (tanishkmangal3@gmail.com).