Magic Resume

The open-source AI resume workbench.
The AI proposes edits in place; you accept or skip each one.

Get Started »  ·  简体中文  ·  Official Site  ·  Feedback

Magic Resume is an open-source AI resume workbench: the AI proposes edits in place and you accept or skip each one. MIT-licensed core, 19 templates, six modes (create, optimize, analyze, translate, interview, export), and multi-device cloud sync.

It pairs a real-time visual editor with multi-model AI — and goes one step further with a native MCP server that lets AI coding tools (Claude Code, Cursor, Windsurf) read and safely patch your resumes without ever guessing the data shape.

Self-hosted vs Cloud

Same features either way. The difference is where your data lives and who supplies the AI.

	Self-hosted (this repo)	Cloud (magic-resume.cn)
License	MIT, free forever	Free tier + paid plans
Account	Not required	Required
Data storage	Your browser's IndexedDB only	Encrypted cloud storage, exportable anytime
Multi-device sync	Not available	On by default
Version history	Local snapshots	Kept in the cloud, restorable
AI	Bring your own API key	Built in, works out of the box
Backend	No backend or database needed	Hosted for you
Templates	19, identical	19, identical

Note

The self-hosted build runs entirely in your browser; resume data never leaves your machine. The cloud build syncs across devices by default, with encrypted transport and storage, and a full export available at any time.

---

Features

Build — visual editing that just works

• Real-time preview — see every change the instant you type.
• 19 professional templates — ATS-friendly, with full color, font, and layout control.
• Deep customization — 22+ font styles, spacing, and layout controls.
• Version history — snapshot and restore any previous version.
• Sharing — share via a unique link with Viewer / Commenter / Editor permissions.

Analyze — Lighthouse-style resume health

• Overall score — a single, honest rating of your resume's impact.
• Detailed breakdown — keyword matching, actionability, and readability.
• Actionable fixes — specific, targeted suggestions, not vague advice.

Optimize — tailor to any job

• JD matching — AI reads a job description and rewrites content to fit.
• Bring your own model — OpenAI, Google Gemini, Anthropic Claude, DeepSeek, or any OpenAI-compatible API.

AI Lab — beyond editing

• Mock interviews — practice with AI feedback on your answers.
• One-click translation — translate a resume into any language, formatting preserved.

Privacy — your data stays yours

• Local-first — stored in IndexedDB by default, no account needed.
• Optional cloud sync — securely sync across devices only if you choose.
• Export anywhere — high-quality PDF or structured JSON.

---

Templates

Nineteen hand-crafted, ATS-friendly templates, each fully customizable (color, font, spacing, layout):

classic · azurill · bronzor · chikorita · ditto · gengar · orange-modern · clean-minimal · teal-professional · red-accent · golden-elegant · product-ops-focus · compact-cn-photo · executive-band · timeline-pro · skills-first · serif-minimal · slate-sidebar · cn-formal-photo

---

MCP Integration

Magic Resume ships a native Model Context Protocol (MCP) server (@magic-resume/mcp) — a first-class integration for AI coding tools.

Configure it once with a personal access token, and your AI assistant can read and patch resumes safely, without touching the raw database:

npx -y @magic-resume/mcp config set --api-url "https://your-api.example.com/api" --pat "mr_pat_xxx"
claude mcp add magic-resume -- npx -y @magic-resume/mcp mcp

Tool	Description
list_resumes	List all resumes in your account
get_resume	Fetch a resume's full content
get_resume_schema	Get the JSON schema for validation
get_resume_editing_guide	AI-readable guide on how to edit resumes
preview_resume_patch	Preview a JSON Patch without applying it
update_resume_content	Apply a JSON Patch to update resume content

Tip

The server is schema-aware and patch-based — AI tools make surgical edits rather than risky full rewrites.

---

Tech Stack

Layer	Technology
Monorepo	Turborepo, pnpm
Framework	Next.js 15 (App Router)
Language	TypeScript
AI / LLM	LangChain, LangGraph, Google GenAI, Anthropic
Auth	Clerk (cloud mode only)
Styling	Tailwind CSS 4
UI	Radix UI, Lucide
Animation	Framer Motion, GSAP
State	Zustand
Storage	IndexedDB
Editor	Tiptap, Monaco
MCP	@modelcontextprotocol/sdk
Schema	Zod
i18n	i18next

---

Quick Start

Magic Resume runs fully in the browser — no backend, no database, no account.

git clone https://github.com/Magic-Resume/Magic-Resume.git
cd Magic-Resume
cp apps/web/.env.example apps/web/.env.local   # set NEXT_PUBLIC_APP_MODE=self-hosted
pnpm install
pnpm run dev

Open http://localhost:3000 and start building.

---

Self-Hosting

Self-hosted mode needs nothing but the browser — set NEXT_PUBLIC_APP_MODE=self-hosted (or leave it unset) and all data persists in IndexedDB. The open-source self-hosted build sends no product analytics.

Cloud mode (auth, cloud sync, sharing) deploys in one click:

Tip

For cloud mode, set NEXT_PUBLIC_APP_MODE=cloud, NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY, and CLERK_SECRET_KEY in your Vercel dashboard.

Environment variables

All variables live in apps/web/.env.local (copy from apps/web/.env.example). Values prefixed with NEXT_PUBLIC_ are inlined into the browser bundle — never put secrets behind that prefix.

Variable	Required in	Default	Purpose
NEXT_PUBLIC_APP_MODE	both	auto	self-hosted or cloud. Auto-detected: cloud when NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY is set, else self-hosted.
NEXT_PUBLIC_APP_URL	both	https://magic-resume.cn	Canonical base URL for OG tags / SEO. Set to your origin (e.g. http://localhost:3000 in dev).
NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY	cloud	—	Clerk publishable key (pk_...).
CLERK_SECRET_KEY	cloud	—	Clerk secret key (sk_...).
NEXT_PUBLIC_CLOUD_API_URL	cloud	http://localhost:3111	NestJS Core API — resumes, settings, sharing, PATs.
BACKEND_URL	AI features	http://localhost:8000	Agent server — interview, translate, AI optimize/analyze.

• self-hosted — pure browser. No Clerk, no Core API; only NEXT_PUBLIC_APP_URL matters.
• cloud — requires the two Clerk keys plus NEXT_PUBLIC_CLOUD_API_URL; AI features additionally require BACKEND_URL.

---

Monorepo

This repository is a Turborepo + pnpm monorepo.

Package	Description
apps/web	Next.js frontend — editor, dashboard, AI Lab
packages/mcp	@magic-resume/mcp — stdio MCP server and CLI
packages/resume-schema	Shared Zod schemas, types, and sample data
packages/resume-templates	Template DSL, renderer, and registry

pnpm install            # install everything
pnpm run dev            # start all workspaces
pnpm run lint           # lint
pnpm run test           # test

# target a single workspace
pnpm --filter @magic-resume/web dev
pnpm --filter @magic-resume/mcp build

---

Contributing

Contributions of every size are welcome — bug fixes, new templates, docs, translations, or features.

Please read the Contributing Guide for local setup, the branch workflow, and the enforced gitmoji + Conventional Commits convention. For bugs or ideas, feel free to file an issue first.

---

Star History

---

Copyright © 2026 Magic Resume Team · MIT licensed.