jobkit-mcp

MCP server for managing job application materials. Reads your resume from the workspace, fetches job postings and company info, compiles LaTeX resumes, and persists research, strategies, and interview prep per company. The host Claude does all reasoning — this server handles data access and persistence.

Quickstart

Requires Python 3.14+ and uv.

git clone <repo-url>
cd jobkit-mcp
uv sync

Add to ~/.claude/settings.json:

{
  "mcpServers": {
    "people": {
      "command": "uv",
      "args": ["run", "--directory", "/path/to/jobkit-mcp", "jobkit-mcp"],
      "env": {
        "JOBKIT_WORKSPACE": "/Users/you/Projects/resume"
      }
    }
  }
}

Tools

Tool	Description
get_profile	Read resume.tex from workspace root as professional profile
list_applications	List all application directories with their files
get_application	Get all text file content for one company
read_application_file	Read any file from an application by filename
save_application_file	Save/update any file in an application
delete_application	Remove an entire application directory
fetch_url	Fetch a URL, return text with HTML stripped
save_job_posting	Fetch URL or accept pasted text, save as job posting
compile_resume	Compile resume.tex with lualatex (company dir or root)
mock_interview	Start a mock interview — returns briefing for Claude to play interviewer

Workspace Layout

$JOBKIT_WORKSPACE/
├── resume.tex              # Base resume (used by get_profile)
├── coverletter.txt         # Base cover letter
├── META.md                 # Base strategy
├── stripe/
│   ├── resume.tex          # Tailored resume
│   ├── coverletter.txt     # Tailored cover letter
│   ├── META.md             # Application strategy
│   └── job_posting.md      # Saved job posting
└── anthropic/
    └── ...

Any file can be stored in a company directory — there is no whitelist. Build artifacts (.aux, .log, .out, .pdf, .gz, .fls, .fdb_latexmk) are skipped in listings.

Documentation

• Tutorial — Walk through a complete application workflow end-to-end
• How-To Guides — Focused recipes for specific tasks
• Reference — Complete tool parameters, return values, and internals
• Explanation — Architecture decisions and design rationale