ApplyGenie

io.github.Eternalcodertanishq3由 GitHub 帳號 Eternalcodertanishq3 發布v0.1.3更新於 Oct 24, 2026

Autonomous job application copilot, ATS form filler, and AI mock interview engine.

已驗證STDIO僅桌面效率與工作流瀏覽器自動化開發者工具

概覽

AI 產生的概覽

一個本機桌面 MCP 伺服器，可在瀏覽器中填寫求職申請表、追蹤申請進度並進行模擬面試練習。

可以做什麼
ApplyGenie 提供約 45 個工具，涵蓋五個面向：桌面控制與螢幕截圖、滑鼠鍵盤輸入以及 Windows 視窗管理；本機候選人設定檔與履歷解析；以 Playwright 為基礎的瀏覽器自動化，可辨識表單欄位、用設定檔內容填寫並上傳履歷；本機 SQLite 申請追蹤器，支援狀態更新與統計；以及模擬面試引擎，包含程式題提示、STAR 回答分析、系統設計階段與評分卡。
適用情境
如果你希望助理協助處理重複的求職表單、在本機記錄投遞情況，並練習程式、行為或系統設計面試，可以考慮加入。它僅限桌面環境，鎖定 Windows 風格的桌面自動化，而非無頭伺服器情境。
執行需求
透過 stdio 在本機執行，從 PyPI 安裝 applygenie-mcp，例如使用 uvx 或 pip。完整的瀏覽器自動化需要 browser 附加元件並安裝 Playwright Chromium。未宣告任何帳號、API 金鑰或環境變數。需要具備顯示器的桌面工作階段。
安裝前請注意
它可以控制滑鼠、鍵盤與視窗，並將設定檔資料輸入網頁表單，因此送出前應檢查填寫內容；表單送出需要明確確認。設定檔資料、履歷快取與申請紀錄存放在使用者家目錄下的本機位置。僅支援桌面，無法在網頁中執行。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 ApplyGenie，將其新增到工作區。
  2. 為需要使用其工具的對話啟用該服務。

Desktop only，透過 STDIO。 STDIO 服務會啟動本機處理程序，因此需要 SourceWeft 桌面主機。

其他 MCP 客戶端

參照 儲存庫 中的啟動說明。

README

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.

[Python 3.10+] [MCP Protocol] [License: MIT] [PyPI] [Total Downloads] [Monthly Downloads] [ApplyGenie on MCPRush] [M8ven Score] [Changelog]

🌟 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)

bash
uvx applygenie-mcp

Option 2: Install via pip

bash
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):

json
{  "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:

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

ToolDescription
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

ToolDescription
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

ToolDescription
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

ToolDescription
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

ToolDescription
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

bash
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 ([email protected]).

來源：README.md，提交 0fd31e9

工具

0
SourceWeft 尚未在沙箱中啟動過該軟體包，因此其工具未知。

版本歷史

2
  1. v0.1.3最新Sep 19, 2026
  2. v0.1.1Sep 17, 2026

評分

已滿足適用檢查項的 50%

該評分衡量此收錄條目的完整程度，依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價，也不表示伺服器安全或目前可用：可連通性是帶日期的檢查結果，評分也不包含安裝量或使用者評價。

  • 近 30 天內可連通25 分

    請保持端點能夠回應 MCP 握手；SourceWeft 每 30 天探測一次。

  • 已宣告認證方式10 分

    請在註冊中心條目中宣告憑證請求標頭，或保持端點能夠回應探測，以便判定其認證方式。

  • 已列出工具15 分

    請讓伺服器在無憑證時回應 tools/list，或發布可在沙箱中無憑證啟動的軟體包。

  • 描述必需15 分
  • 圖示10 分
  • 文件15 分
  • 授權條款10 分