mcp-resume

MCP server for resume and cover letter writing, CV documents in Word format. Resumes and cover letters as Word files from one profile, never invented.

Works with Claude Desktop, Claude Code, Cursor and any Model Context Protocol client. Runs on your own machine, or hosted with no install.

Product page: https://mcp.zovo.one/s/resume — what it does, the tools it exposes, and a live token endpoint.

Install

Hosted, nothing to install. Get a token from https://mcp.zovo.one/mcp/connect (the connect page) or https://mcp.zovo.one/mcp/token (the same token as JSON); a free anonymous one is issued on the spot and a Pro key works the same way. Then point an MCP client at https://mcp.zovo.one/mcp/resume over streamable-http and send the token as Authorization: Bearer <token>.

If your client cannot set headers, put the token in the path instead: https://mcp.zovo.one/mcp/resume/t/<token>. Both forms work. The bare URL with no token answers 401 on tools/call, so the token is not optional.

Claude Desktop, one click. Download resume.mcpb from the latest release and double-click it.

From source. The mirror is self-contained: every @theluckystrike/* dependency is vendored, so a fresh clone builds with no extra setup.

git clone https://github.com/theluckystrike/mcp-resume.git
cd mcp-resume
npm install && npm run build

Then point your client at the built entry point:

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

@theluckystrike/mcp-resume is not published on npm yet, so an npx -y @theluckystrike/mcp-resume command will fail. The three paths above are the working ones and each is exercised by CI.

Read-only mirror of mcp-servers/servers/resume. See MIRROR.md.

Store your CV facts once. Then say "tailor my resume to this posting and write the cover letter" and get two real .docx files: a resume whose bullets are reordered by relevance to the posting, trimmed to fit the page count you asked for, with the posting's keywords in bold where you actually have them, and a one-page cover letter that states nothing you did not put in your profile. Where a fact is missing, the letter leaves a bracketed prompt like [add: metric] instead of inventing a number. It also reads an existing resume .docx back into the profile shape, exports markdown for an ATS box, and writes printable HTML. Everything runs locally: no upload, no account, no native dependency.

In the official MCP Registry (io.github.theluckystrike/resume-cover-letter-docx-generator).

Listed on the AI Product Index — live remote endpoint at mcp.zovo.one/s/resume, free tier, no signup.

60-second install

npm publish for @theluckystrike/mcp-resume is pending. Until then, the .mcpb one-click bundle or a clone+build is the working path, both are verified below.

One-click (.mcpb): download resume.mcpb from the latest release and double-click it in Claude Desktop: https://github.com/theluckystrike/mcp-servers/releases/latest

(claude_desktop_config.json):

{
  "mcpServers": {
    "resume": {
      "command": "npx",
      "args": ["-y", "@theluckystrike/mcp-resume"]
    }
  }
}

Claude Code:

claude mcp add resume -- npx -y @theluckystrike/mcp-resume

(.cursor/mcp.json):

{
  "mcpServers": {
    "resume": {
      "command": "npx",
      "args": ["-y", "@theluckystrike/mcp-resume"]
    }
  }
}

The npx form above starts working the moment the package is published. Until then, use the .mcpb bundle above, or build from source with exactly these three commands:

git clone https://github.com/theluckystrike/mcp-servers.git && cd mcp-servers
npm install
npm run build -w packages/mcp-license -w servers/docx -w servers/resume

Then point your client's command at node with one arg: the absolute path to servers/resume/dist/index.js. servers/docx is in that build list because this server shares the document engine with mcp-docx rather than carrying a second copy of it.

To run in Pro mode set MCP_LICENSE_KEY in the same config block, or call license_activate once with your key.

Tools

Resource: resume://profile returns the stored profile as JSON. Prompt: apply_to_job chains tailor_to_job, resume_create and cover_letter_create against one posting, and makes the model report the gaps back to you instead of filling them in.

The cover letter does not invent facts

This is the part that matters, so it is enforced rather than requested:

• Every proof line in the letter is a verbatim bullet from your profile. The letter never paraphrases a bullet into a stronger claim.
• A highlights entry you pass is checked against the profile first. If the profile does not support it, it is printed as [add: "..." is not in your profile - add it there or drop it], not as a claim.
• Before the file is written, every digit run in the letter is checked against your profile and the arguments you passed. A number that traces to neither is a refusal, not a warning, the tool returns an error and writes nothing. The job description is deliberately not an allowed source: the employer's revenue, headcount and throughput figures are theirs, and the letter will never restate one as yours. Comparison is on whole numbers, so a profile holding 2012 does not license a letter claiming 12.
• Each proof bullet is printed under the role it actually belongs to. A letter that quotes work from two employers prints two headings; work from an earlier job is never listed under your current one.
• A bullet that carries no figure gets [add: metric] appended, so the letter asks you for the result instead of guessing one.
• Missing keywords from tailor_to_job never become skills. The response says so in plain words.

What you can say

Worked example

You: Tailor my resume to this posting and write the letter. One page.

  tailor_to_job { job_description: "...payments platform, PostgreSQL, Kubernetes..." }
  -> matched: payments, postgresql, kubernetes   missing: rust
  -> coverage 75%

  resume_create { style: "modern", target_role: "Staff Engineer",
                  keywords: ["postgresql", "kubernetes", "rust"], max_pages: 1 }
  -> estimated_pages 1, words_used 402 of a 384-word budget for bullets
  -> bullets_dropped: ["Wrote the deployment pipeline"]
  -> keywords_missing: ["rust"]  (not added anywhere)
  -> ~/.local/share/mcp-servers/resume/documents/ada-rowe-resume.docx

  cover_letter_create { company: "Zeta Systems", role: "Staff Engineer", tone: "formal" }
  -> fills_required: ["[add: metric]"]
  -> ~/.local/share/mcp-servers/resume/documents/zeta-systems-staff-engineer-cover-letter.docx

How page trimming works

There is no page-layout engine in pure JavaScript, so this server does not pretend to measure one. It uses a word budget instead: a full A4 page at 11pt Calibri with 2cm margins holds about 520 words of body text, and a resume spends roughly a seventh of that page on headings, blank lines and the contact block. The net figure is 450 words per page (540 for the compact style). The contact block, summary, skills, role headers, education, certifications and languages are counted first; whatever budget is left goes to experience bullets, highest score first. Every role keeps its first bullet before any role gets a second one, so trimming never leaves a job on the page with nothing under it. The response reports the budget, the words used, the estimated page count and every bullet it dropped, so you can see the decision rather than discover it in Word.

Recency is part of that score, and it is read off array order, not off start/end text. profile_set enforces the ordering that makes that safe: however you list roles when you call it, the stored profile always ends up newest-first, an open role with no end first, then by end descending, then by start descending. Enter roles in any order; the stored order (and the cover-letter bullet ranking in cover_letter_create) is always correct.

Free vs Pro

Get Pro: https://mcp.zovo.one/buy/resume ($19 one-time, or $39 for the whole bundle).

Privacy

All data stays local. Your CV never leaves the machine: the server reads and writes files on your computer, stores the profile under your data directory, and makes no network request of any kind, not for licensing (keys are verified offline), not for fonts, not for telemetry.

Pairs with

• mcp-docx, this server imports its document engine (@theluckystrike/mcp-docx/lib); install docx too when you also want proposals, contracts and markdown-to-Word.
• office-suite, several servers behind one install, one config entry.

Troubleshooting

• npx hangs or fails to find the package: npm publish for this package is pending. Use the .mcpb bundle or the clone-and-build path above until it lands.
• "no profile stored": run profile_set once, or resume_read {path, save: true} from an existing resume.
• The resume dropped a bullet I wanted: raise max_pages, or pass the posting's words in keywords so the bullet outranks the others. bullets_dropped in the response names every one that did not fit.
• Short skill names: tailor_to_job keeps two- and one-character skills, Go, C, R, C#, C++, F#, Qt, UI, UX, QA, ML, AI, AR, VR, JS, TS, K8s, AWS, GCP, SQL, iOS, plus any short word that appears in your own skills list, and ranks a known skill above a longer word of the same frequency.
• A keyword I have is reported missing: matching is on word boundaries, so go does not match Google and k8s does not match Kubernetes. Add the exact word to your skills if it is true.
• resume_read put a role in unparsed: resumes have no schema. Nothing is dropped silently, fix the fields and pass them to profile_set.
• There is no resume_to_pdf: every pure-JavaScript route from Word to PDF needs a native dependency or a cloud API. resume_to_html writes semantic HTML with a print stylesheet; print that to PDF.
• Node version: requires Node >= 18. Check with node -v.

MIT licensed. Support: support@zovo.one

Built by theluckystrike.

One business profile for the whole suite

Your identity is stored once, at ${XDG_DATA_HOME:-~/.local/share}/mcp-servers/profile/business.json, and every server in the suite reads it: the invoice issuer, the docx letterhead, the recurring issuer, expense-tracker's default VAT rate, time-tracker's and timezone's home zone, and the resume and contract letterheads. Set it once with business_set (invoice or docx) - you never repeat it anywhere else. An email address is only ever taken from that profile or from an explicit argument; when none is stored, documents show [add: email] and the tool says so rather than letting anyone improvise an address.

Use these docs as an MCP server

Any MCP client (Claude, Cursor, Windsurf, VS Code) can read this repository's documentation directly via GitMCP — no install:

• Docs MCP URL: https://gitmcp.io/theluckystrike/mcp-resume