convez.ai
MCP ServerProdutividadeIntermediário
ahardy42

Job Search MCP Server

ahardy42/job-search-mcp

Automação completa de busca de emprego e pesquisa corporativa via MCP

Ver no GitHub
InicialPluginScore · de 0 a 100

Servidor MCP que integra busca de vagas no LinkedIn, análise de perfil profissional e pesquisa de empresas. Fornece ferramentas para buscar posições com filtros avançados, carregar dados de perfil, obter descrições completas de vagas e pesquisar informações sobre empresas com web scraping.

Instalação

$npx convez-ai ahardy42/job-search-mcp

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
28%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Buscar vagas no LinkedIn com filtros de senioridade, localização e tipo de contrato
  • Analisar alinhamento entre perfil profissional e descrições de vagas
  • Pesquisar empresas e coletar informações de Glassdoor e sites corporativos
  • Gerar cartas de apresentação baseadas em dados de perfil e descrições de vagas

Veredito

MCP server robusto e bem-estruturado para automação de processos de busca de emprego. Requer Node.js 20+ e configuração manual de perfil, mas oferece funcionalidades avançadas de pesquisa e análise com batching inteligente para evitar rate limiting.

README

A tradução para o português deste README ainda está na fila.

job-search-mcp-server

A stdio MCP server for job searching, profile loading, and company research. Exposes 4 tools to any MCP-compatible client.

Tools

job_search

Search LinkedIn for job listings with filters for keywords, location, job type, remote preference, experience level, salary, and recency.

Defaults are tuned for senior, remote, full-time roles posted in the past week.

profile_data

Load and return all profile sections (experience, skills, education, etc.) and the full concatenated text. The calling LLM uses this data directly for job alignment analysis, cover letter writing, or any other profile-based task.

job_description_search

Fetch full job descriptions from LinkedIn job URLs. Accepts a list of URLs (from job_search results) and returns the complete description text for each. Fetches run in parallel batches with delays to avoid rate limiting.

company_research

Research a company using DuckDuckGo web search (via Playwright). Searches for company info, recent news, and Glassdoor reviews. Optionally scrapes the company's own website. Returns structured data for "what they say vs what they actually do" analysis.

Installation

git clone https://github.com/ahardy42/job-search-mcp.git && cd job-search-mcp
npm install
npm run build
npx playwright install chromium  # required for company_research and job_description_search

Requires Node.js >= 20.

To verify everything works:

npm test

Then connect the server to your MCP client of choice (see Connecting to Your MCP Client below).

Profile Setup

The profile_data tool requires a profile directory with your experience data.

  1. Create a profile/ directory in the project root (it's gitignored)
  2. Add your files (Markdown, PDF, or DOCX)
  3. Create profile/manifest.json listing your files:
{
  "sections": [
    { "id": "summary", "label": "Professional Summary", "source": "summary.md", "format": "md", "order": 1 },
    { "id": "experience", "label": "Work Experience", "source": "experience.pdf", "format": "pdf", "order": 2 },
    { "id": "skills", "label": "Technical Skills", "source": "skills.md", "format": "md", "order": 3 }
  ]
}

Supported formats: md, txt, pdf, docx.

The profile is loaded once and cached for the server session. Files are processed in explicit order for deterministic output.

Override the profile directory path via the PROFILE_DIR environment variable.

Connecting to Your MCP Client

This is a stdio server — it works with any MCP-compatible client. Replace /absolute/path/to/job-search-mcp/dist/index.js with the actual path on your machine.

Node version managers (nvm, nodenv, asdf): MCP clients spawn the server as a subprocess and may not inherit your shell's PATH. Use the full path to your node binary (run which node to find it).

Claude Code (CLI)

Via command line:

claude mcp add --transport stdio job-search -- node /absolute/path/to/job-search-mcp/dist/index.js

Or manually in .mcp.json (project scope, checked into version control):

{
  "mcpServers": {
    "job-search": {
      "command": "node",
      "args": ["/absolute/path/to/job-search-mcp/dist/index.js"]
    }
  }
}

Restart Claude Code or run /mcp to verify the server is connected.

Claude Desktop

Edit the config file:

  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
  • Windows: %APPDATA%\Claude\claude_desktop_config.json
{
  "mcpServers": {
    "job-search": {
      "command": "node",
      "args": ["/absolute/path/to/job-search-mcp/dist/index.js"]
    }
  }
}

Quit and restart Claude Desktop completely (not just the window). Look for the hammer icon in the chat input.

VS Code

Create .vscode/mcp.json in your workspace:

{
  "mcpServers": {
    "job-search": {
      "type": "stdio",
      "command": "node",
      "args": ["/absolute/path/to/job-search-mcp/dist/index.js"]
    }
  }
}

Or via command line:

code --add-mcp '{"name":"job-search","type":"stdio","command":"node","args":["/absolute/path/to/job-search-mcp/dist/index.js"]}'

Other MCP Clients (Cursor, Codeium, etc.)

Most clients use the same standard format:

{
  "mcpServers": {
    "job-search": {
      "type": "stdio",
      "command": "node",
      "args": ["/absolute/path/to/job-search-mcp/dist/index.js"]
    }
  }
}

Environment Variables

Pass PROFILE_DIR to override the default profile location:

{
  "mcpServers": {
    "job-search": {
      "command": "node",
      "args": ["/absolute/path/to/job-search-mcp/dist/index.js"],
      "env": {
        "PROFILE_DIR": "/path/to/your/profile"
      }
    }
  }
}

Development

npm run dev    # watch mode with tsx
npm run build  # compile TypeScript
npm start      # run compiled server

Testing

The project uses Vitest with 102 tests across unit, tool, and integration layers.

npm test          # run all tests once
npm run test:watch  # re-run on file changes

What's covered:

  • Service unit tests — profile loading/caching, LinkedIn HTML parsing and URL building, SSRF mitigations (private IP rejection, DNS rebinding, redirect validation), HTTPS enforcement
  • Tool unit tests — Zod schema validation, handler responses, CHARACTER_LIMIT truncation, error handling for each of the 4 tools
  • Integration tests — full MCP client-server round trips using in-memory transport, tool listing, end-to-end tool calls, and input validation at the protocol level

Security-focused tests verify that path traversal, SSRF protections, and input validation don't regress.

MCP Inspector

The MCP Inspector provides an interactive web UI for testing tool calls against the running server.

npm run inspect      # against compiled dist/ (run npm run build first)
npm run inspect:dev  # against src/ via tsx (no build needed)

Opens at http://localhost:6274 where you can browse tools, edit inputs, and see raw responses.

Ver o original no GitHub

Bom para

Relacionados