Resuml + Claude Code: tailoring a resume from a chat

A quick follow-up to Introducing Resuml. The CLI is still there, the YAML is still there, but the part I’m most excited about is that resuml now ships a Model Context Protocol server. That means Claude Code (or any MCP client, really) can drive the whole thing through a chat.

Setup, in one line

In the project where your resume.yaml lives:

claude mcp add resuml -- npx resuml mcp

Add --scope user if you want it everywhere, or commit a .mcp.json if you’re sharing the setup with a team:

{
  "mcpServers": {
    "resuml": {
      "command": "npx",
      "args": ["resuml", "mcp"]
    }
  }
}

That’s it. claude mcp list will confirm it’s wired up.

What changes once it’s wired up

Honestly, the workflow flips. Before, I’d open the YAML, eyeball a job description, rewrite a few bullets, run resuml validate --ats --jd job.txt, scroll through the output, edit again. Now I save the job ad as jd.txt and say something like:

Tailor resume.yaml to jd.txt, iterate until the ATS score is at least 80, then render with the stackoverflow theme and export a PDF.

Claude Code reads the YAML, runs the ATS check, sees which keywords are missing and which bullets are weak, edits the file, re-checks, and keeps going until the score clears. Then it renders the HTML and exports the PDF. I just watch.

The reason this works at all is that resuml’s ATS scoring is deterministic and offline. No API keys, no LLM grading itself. That matters a lot here, because it means the agent has an honest feedback signal it can’t talk its way around. To raise the score, it has to actually improve the resume: add the missing skills you genuinely have, rewrite passive bullets into action verbs, quantify impact, fix date gaps. With a job description, the score is 60% generic checks and 40% keyword match.

The full toolbox

For reference, the MCP server exposes:

• resuml_init_resume to scaffold a starter YAML
• resuml_validate against the JSON Resume schema
• resuml_ats_check for the deterministic score and keyword matching
• resuml_render for HTML with any jsonresume-theme-*
• resuml_list_themes to see what’s installed
• resuml_export_pdf for PDF via Playwright

It also exposes the schema, the ATS rubric, and the theme catalog as resources, plus a few prompts (tailor-resume-to-jd, optimize-ats-score, review-resume) if you’d rather start from a template than write the request yourself.

Why I think it’s worth your five minutes

Most “AI resume” tools either rewrite your resume into something generic or lock you into a proprietary editor. This is the opposite: your resume stays as plain YAML in your repo, you keep version control, and the agent just helps you iterate faster against a deterministic rubric. If you already use Claude Code, you’re maybe 30 seconds of setup away from trying it.

Repo is at github.com/phoinixi/resuml. The hosted web app, with 300+ themes pre-installed, is at resuml.app.