cv-mirror-mcp
MCP server that simulates how 5 real ATS parsers (Workday, Greenhouse, Lever, Taleo, iCIMS) read a CV. Pure JS lint engine, MIT licensed, no network calls in lint logic. Vendor-specific rules cite their public sources.
README
cv-mirror-mcp
Model Context Protocol server that lints a CV against 5 real ATS parsers.
What it does
Exposes three MCP tools so any MCP-compatible AI agent (Claude Code, Cursor, Windsurf, Zed, OpenCode, Cline, etc.) can analyse a CV against the documented quirks of:
- Workday
- Greenhouse
- Lever
- Taleo (Oracle)
- iCIMS
These are the 5 most-used enterprise ATS systems. Each one parses CVs slightly differently. A "0–100 ATS score" averages those differences into noise. This server surfaces what each one actually does — column handling, emoji stripping, header-footer dropping, date-format quirks — as concrete lint findings with concrete fixes.
The lint rules are derived from public vendor documentation. See docs/vendor-sources.md for citations.
Tools
| Tool | Description |
|---|---|
analyze_cv |
Full report. Takes a CV file path (PDF or DOCX), returns per-vendor lint findings with severity (error / warn / info) and fixes. |
lint_for_vendor |
Single-vendor lint. Takes a path + vendor name. Faster when the user only cares about one ATS. |
get_express_url |
Returns the URL of the visual web companion (cv-mirror-web.vercel.app) for users who prefer a side-by-side parser view in their browser instead of an agent session. |
Install
npm (recommended)
npm install -g cv-mirror-mcp
MCP client config
Add to your MCP client's config file:
Claude Code (~/.claude/mcp.json or .claude.json in project):
{
"mcpServers": {
"cv-mirror": {
"command": "npx",
"args": ["-y", "cv-mirror-mcp"]
}
}
}
Cursor / Windsurf / Zed / Cline: same config format (paste into wherever your client reads mcpServers).
Restart the client. The three tools above auto-register.
Usage
Once installed, just ask your agent:
Scan my CV at
~/Documents/resume.pdfand tell me what each ATS would do to it.
The agent calls analyze_cv, the lint engine runs locally (no network, no upload), and the agent surfaces the structured output:
CV Mirror — multi-vendor ATS lint report
Source: /Users/jane/Documents/resume.pdf
Format: pdf
Pages: 2
Words: 542
[ERROR] Workday
- ERROR WORKDAY_MULTI_COLUMN: 35% of lines look multi-column. Workday's parser
reads left-to-right and interleaves both columns into one stream.
Fix: Convert to single-column layout. Move sidebars (Skills, Tools, Languages)
above or below the main content.
[OK] Greenhouse
No issues detected by the simulated parser.
[WARN] Lever
- WARN LEVER_HEADER_FOOTER: Header/footer-like text detected ("Page 1 of 2").
Lever historically drops content placed in PDF headers/footers.
Fix: Remove headers and footers. Page numbers are not needed on a CV.
...
What the lint actually checks
These are real heuristics derived from public ATS docs and vendor support articles. The full list with citations lives in docs/vendor-sources.md.
| Rule | Vendor | Severity | Trigger |
|---|---|---|---|
WORKDAY_MULTI_COLUMN |
Workday | error | >15% of lines have a 5+-space gap |
WORKDAY_QUARTER_DATES |
Workday | warn | Dates use "Q3 2024" format |
WORKDAY_CONTACT_MISSING |
Workday | error | No email or phone in plain text |
GREENHOUSE_EMOJI |
Greenhouse | warn | Any emoji codepoint detected |
GREENHOUSE_FANCY_BULLETS |
Greenhouse | info | Non-standard bullet glyphs |
GREENHOUSE_NONSTANDARD_HEADER |
Greenhouse | warn | "My Story", "Highlights Reel", etc. |
LEVER_HEADER_FOOTER |
Lever | warn | "Page X of Y" pattern detected |
LEVER_NO_STANDARD_HEADERS |
Lever | error | No "Experience"/"Education" sections |
TALEO_ISO_DATES |
Taleo | warn | ISO dates outnumber Month-Year dates |
TALEO_LOW_TEXT_DENSITY |
Taleo | error | <1 word per kB of file size |
TALEO_SMART_QUOTES |
Taleo | info | Curly quotes detected |
ICIMS_MULTI_COLUMN |
iCIMS | error | >20% of lines multi-column |
ICIMS_FEW_HEADERS |
iCIMS | warn | <2 standard headers found |
Requirements
- Node.js >= 18
- An MCP-compatible client (Claude Code, Cursor, Windsurf, Zed, Cline, etc.)
Privacy
The server runs locally inside your agent's process. No upload endpoint. No telemetry. Your CV bytes never leave your machine.
The visual web companion at cv-mirror-web.vercel.app is also fully client-side — same engine, runs in the browser.
Tests
git clone https://github.com/goofypluto999/cv-mirror-mcp.git
cd cv-mirror-mcp
npm install
npm test
19 unit tests covering signal extraction and per-vendor rule firing.
Sister projects
- CV Mirror (web) — visual side-by-side parser view. Same lint engine. Drag-drop a PDF, see the reading-order overlay on the actual document.
- Vantage AI — once your CV passes the parsers, the next problem is the application: tailored cover letter, mock interview, fit analysis. Vantage handles that flow. Paid (£5 starter / 20 tokens, 10 free tokens at signup).
Contributing
Pull requests welcome. Particularly interested in:
- More vendor simulators (BambooHR, SmartRecruiters, JazzHR, Recruitee, etc.)
- Updated rules when vendors change their parsing behaviour (open an issue with the source link)
- Translations of the lint output
License
MIT. Workday, Greenhouse, Lever, Taleo, and iCIMS are trademarks of their respective owners. This project is not affiliated with any of them; the names are used for descriptive reference (nominative fair use).
Built by Vantage Labs.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。