Lathe
Local MCP server that shapes tailored resume PDFs with live browser preview.
README
Lathe
Local MCP server that shapes tailored resume PDFs with a live browser preview.
Lathe is a small Model Context Protocol server you run on your own machine. You talk to Claude about your background, your target job, and the kind of resume you want; Claude shapes the content; Lathe renders it in a real browser window that updates as you iterate. When you're happy, Lathe exports a print-ready PDF.
Nothing about you leaves your laptop.
Why this exists
A lot of resume MCPs ship your data to a third-party API to render the PDF. The most popular one (resumake-mcp) depends on an upstream service that's been shut down, and pins a vulnerable version of the MCP SDK. Lathe takes the opposite approach: render everything locally with headless Chromium, no external services, no accounts.
Features
- Live preview — A Chromium window opens on first use and stays put. Every edit Claude pushes refreshes the page in place, so you watch the resume change as you give feedback.
- HTML/CSS templates — Three starter templates (
modern,classic,minimal). Hand-write your own, or describe one in chat and let Claude generate it. - Print-ready PDF —
letterpaper, real CSS, real fonts. Exported on demand. - Local-first — No third-party APIs. No accounts. Your data stays on your machine.
- Strict TypeScript —
strict: true,noUncheckedIndexedAccess: true. - Zero secrets — no env vars, no keys, no auth.
How it works
You ────▶ Claude ────▶ Lathe MCP ──▶ Handlebars render ──▶ work/current.html
│
▼
Puppeteer Chromium
(visible window)
│
────▶ page.reload() ◀──┘
│
▼
You see it change
│
▼
"Export it"
│
▼
output/*.pdf
Install
git clone https://github.com/itonskie/lathe.git
cd lathe
npm install # also downloads a Chromium for Puppeteer (~170 MB on first install)
npm run build
Then add Lathe to your Claude Desktop config:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
Linux: ~/.config/Claude/claude_desktop_config.json
Recommended (works with nvm / fnm / system node):
{
"mcpServers": {
"lathe": {
"command": "/absolute/path/to/lathe/bin/lathe.sh",
"args": []
}
}
}
The wrapper script in bin/lathe.sh sources nvm if it's installed, so it survives nvm install upgrades.
System-node alternative (no wrapper):
{
"mcpServers": {
"lathe": {
"command": "node",
"args": ["/absolute/path/to/lathe/dist/server.js"]
}
}
}
Only works if node is on Claude Desktop's PATH — i.e., installed via Homebrew or system package manager, not nvm. (Claude Desktop spawns subprocesses with a minimal PATH that doesn't see ~/.nvm/.)
Fully quit Claude Desktop (⌘Q on macOS — closing the window isn't enough) and reopen. Lathe shows up in the tools list.
Usage
Talk to Claude naturally:
"Here's my background [paste]. I'm applying for this Senior Backend Engineer role at Stripe [paste JD]. Use the modern template, open the preview, and let's iterate."
Claude will:
- Call
lathe_templatesto confirm what's available. - Extract a structured resume from your inputs.
- Call
lathe_preview— a Chromium window opens with your resume. - You give feedback in chat ("make the first Acme bullet stronger", "drop the GPA").
- Claude calls
lathe_shapeafter each round — the window auto-refreshes. - When it looks right, ask for the PDF — Claude calls
lathe_finish.
Tools
| Tool | What it does |
|---|---|
lathe_templates |
List templates in templates/ |
lathe_preview(data, template) |
Render + open the live browser window |
lathe_shape(data, template?) |
Push updates; browser auto-refreshes |
lathe_finish(filename?, folderPath?) |
Export the current preview as PDF |
lathe_add_template(name, html) |
Save a new template |
Templates
Templates are plain HTML files in templates/, with Handlebars placeholders:
<h1>{{basics.name}}</h1>
{{#each work}}
<div>
<strong>{{position}}</strong> at {{company}}
<ul>{{#each highlights}}<li>{{this}}</li>{{/each}}</ul>
</div>
{{/each}}
One custom helper is registered: {{join keywords ", "}} — joins an array with a separator.
Templates assume this data shape, but any field is optional:
{
basics: { name, headline, email, phone, location, website, summary }
work: [{ company, position, location, startDate, endDate, highlights[] }]
education:[{ institution, area, studyType, location, startDate, endDate, gpa }]
skills: [{ name, keywords[] }]
projects: [{ name, description, url, keywords[], highlights[] }]
}
To add a template, drop an HTML file into templates/, or have Claude generate one and save it via lathe_add_template.
Development
npm run dev # tsx watch — restart on save
npm run typecheck # strict tsc check
npm test # smoke test: render all templates + export PDFs (no visible browser)
npm run build # compile to dist/
Project layout
lathe/
├── src/
│ ├── server.ts # MCP wiring (tool registration, transport)
│ └── lathe.ts # Core: render + Puppeteer + PDF
├── templates/ # HTML templates (Handlebars)
├── test/smoke.ts # End-to-end render + PDF test
├── work/ # Rendered HTML lives here (gitignored)
├── output/ # Exported PDFs (gitignored)
└── dist/ # Compiled JS (gitignored)
License
MIT © 2026 itonskie
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.