Magic-Resume
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, cloud sync.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Badge for your README
Markdown
[](/listings/magic-resume--magic-resume)HTML
<a href="/listings/magic-resume--magic-resume"><img src="/api/listings/magic-resume--magic-resume/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
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](http://localhost:3000) and start building.
---
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
README.md · README · 280 lines
View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)
This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.
- <div align="center">
- <a href="https://magic-resume.cn">
- <img width="112" alt="Kenmet" src="./apps/web/public/magic-resume-mark.png">
- </a>
- <h1>Kenmet</h1>
- <p><strong>An open-source resume editor with a self-hosted AI agent.</strong><br/>Your resumes stay in your browser. The AI runs on your own model key.</p>
- <p>
- <a href="#quick-start"><strong>Quick start</strong></a>
- ·
- <a href="./README.zh-CN.md">简体中文</a>
- ·
- <a href="https://magic-resume.cn">magic-resume.cn</a>
- ·
- <a href="https://github.com/Magic-Resume/Magic-Resume/issues">Report an issue</a>
- </p>
- [![][github-stars-shield]][github-stars-link]
- [![][github-forks-shield]][github-forks-link]
- [![][github-contributors-shield]][github-contributors-link]
- [![][github-issues-shield]][github-issues-link]
- [![][github-license-shield]][github-license-link]
- </div>
- <br/>
- 
- **Kenmet (formerly Magic Resume) is an open-source AI job-search workbench: the AI proposes resume edits in place and you accept or skip each one.** MIT-licensed core, 20 templates, six modes (create, optimize, analyze, translate, interview, export), cloud sync.
- > Kenmet (formerly Magic Resume, magic-resume.cn) and the Magic Resume at magicv.art (魔方简历) are two independent projects maintained by different teams.
- This repository is the open-source edition. It contains the editor and **Polaris**, the AI resume assistant, together with the agent service that powers it. There is no account and no resume library on a server: resumes and one conversation per resume live in your browser's IndexedDB, and Polaris calls the model provider you configure with your own key. Interviews and multi-device cloud sync are part of the hosted product at [magic-resume.cn](https://magic-resume.cn).
- <details>
- <summary><strong>Table of contents</strong></summary>
- - [Features](#features)
- - [Quick start](#quick-start)
- - [How Polaris works](#how-polaris-works)
- - [Configuration](#configuration)
- - [Open source and magic-resume.cn](#open-source-and-magic-resumecn)
- - [Repository layout](#repository-layout)
- - [Tech stack](#tech-stack)
- - [MCP](#mcp)
- - [Roadmap](#roadmap)
- - [Contributing](#contributing)
- - [License](#license)
- </details>
- ## Features
- ### Editor
- 
- - **Live preview at print fidelity.** The page re-renders as you type; exports match what you see.
- - **20 templates.** ATS-friendly, each with its own color, font, density, spacing and layout controls, on A4 or Letter.
- - **Flexible sections.** Add, rename, reorder and hide sections; rich text with Tiptap; custom fields.
- - **Import and export.** Open Kenmet JSON offline. Export to PDF, PNG or JSON.
- - **Local-first.** Everything is stored in IndexedDB. Nothing leaves the browser unless you use AI.
- - **Light and dark themes, English and Chinese.**
- ### Polaris
- - **Proposals, not overwrites.** Every change the AI makes appears as a diff on the canvas, with a reason for each edit. You accept or skip changes one by one; nothing is saved until you accept.
- - **Skills.** Create a resume from scratch, optimize it for a job description, analyze it, score its fit against a role, or translate it. Each skill is a plain Markdown file you can read and change.
- - **Three modes.** *Co-create* proposes edits, *Plan* reads the resume and proposes a plan without changing it, and *Ask* answers general questions without seeing your resume at all. The boundary is enforced by the backend, not only by the prompt.
- - **Asks when it needs to.** When a fact is missing, such as the target role, the agent shows a structured form, pauses, and continues once you answer. The pause survives a page reload or a server restart.
- - **Import any resume.** PDF, Word, images, Markdown or plain text. Scanned documents are read by a vision model.
- - **Your own model.** Any OpenAI-compatible API, with presets for common providers. The key is encrypted in the browser and sent only with each request; the server never stores it.
- - **No made-up experience.** The agent checks every proposed resume against your source material and flags employers, dates, credentials or numbers it cannot find there.
- ## Quick start
- ### Docker Compose (recommended)
- Runs the editor and the agent together. Requires Docker with Compose v2.
- ```bash
- git clone https://github.com/Magic-Resume/Magic-Resume.git
- cd Magic-Resume
- docker compose up --build
- ```
- Open [http://localhost:3000](http://localhost:3000), go to **Settings → Model**, choose a provider, enter your API key and model, test the connection and save. Polaris is then ready.
- The editor listens on `127.0.0.1:3000`. The agent has no published port and is reachable only from the editor container; its sessions are kept in the `agent-data` volume.
- ### Local development
- Requires Node.js 22.13 or later and pnpm 10.28.
- ```bash
- git clone https://github.com/Magic-Resume/Magic-Resume.git
- cd Magic-Resume
- pnpm install
- cp apps/web/.env.example apps/web/.env.local # then uncomment AI_BACKEND_URL=http://localhost:4000
- pnpm dev
- ```
- `pnpm dev` starts the editor on [http://localhost:3000](http://localhost:3000) and the agent on `http://127.0.0.1:4000`. Editing works without the agent; only Polaris needs it.
- To point the agent at a model server on your own machine or network (for example a local OpenAI-compatible proxy), start it with `AGENT_ALLOW_PRIVATE_MODEL_HOSTS=true`. This switch is ignored when `NODE_ENV=production`.
- ### Editor only
- The editor is a standard Next.js app and runs without the agent:
- ```bash
- docker build -f apps/web/Dockerfile -t magic-resume-web .
- docker run -p 3000:3000 magic-resume-web
- ```
- [](https://vercel.com/new/clone?repository-url=https%3A%2F%2Fgithub.com%2FMagic-Resume%2FMagic-Resume)
- To enable Polaris later, run the agent somewhere the editor can reach and set `AI_BACKEND_URL` on the editor.
- ## How Polaris works
- ```mermaid
- flowchart LR
- browser["Browser<br/>editor · IndexedDB"] -->|same-origin requests| web["Next.js routes<br/>apps/web"]
- web -->|AI_BACKEND_URL| agent["Agent<br/>apps/agent"]
- agent -->|your key| provider["Your model provider<br/>OpenAI-compatible API"]
- agent --- store[("SQLite<br/>session checkpoints")]
- ```
- - **The browser owns the resume.** Each request carries a snapshot of the current resume (except in Ask mode). The agent works on a draft and returns proposals; the browser saves what you accept. The draft lives only in the session's checkpoint and is deleted with it.
- - **The agent** (`apps/agent`) is a [DeepAgents](https://github.com/langchain-ai/deepagentsjs) harness on LangGraph with a Fastify HTTP layer. It mounts the skills directory read-only, gives the model a scratch workspace holding the resume draft, and runs a middleware chain that enforces the run budget, mode boundaries, each skill's tool allowlist, resume schema validation and the fabrication check.
- - **Sessions** are LangGraph checkpoints in SQLite, so a paused form, a page reload or a restart resumes where it left off. Idle sessions expire after seven days.
- - **Your key** stays in the request: it is never logged, written to a checkpoint or stored on disk. The agent resolves the provider's address once per request, pins the connection to it, refuses private network addresses and does not follow redirects.
- - **The protocol** between the editor and the agent is documented in [docs/reference/ai-backend.md](./docs/reference/ai-backend.md), and the agent's design in [docs/specs/agent-backend/design.md](./docs/specs/agent-backend/design.md). You can replace the agent with your own implementation of the same contract.
- > [!IMPORTANT]
- > The agent has no user accounts. Keep it on a private network, as Docker Compose does, or put authentication in front of it before exposing it to other people.
- ## Configuration
- Model settings (provider, key, base URL, model, output limit, reasoning strength) are configured per browser in **Settings → Model**. Server settings are environment variables; all are optional.
- <details>
- <summary><strong>Editor (<code>apps/web</code>)</strong></summary>
- <br/>
- | Variable | Default | Purpose |
- |---|---|---|
- | `AI_BACKEND_URL` | — | Server-only origin of the agent, without `/api`. AI requests return 503 when it is unset. |
- | `NEXT_PUBLIC_APP_URL` | `http://localhost:3000` | Public origin, used for `robots.txt` and page metadata. |
- | `APP_PUBLIC_URL` | — | Same as above, read at request time so one image can serve several domains. |
- | `NEXT_PUBLIC_DOCS_URL` | — | Target of the "Help center" menu item; hidden when unset. |
- </details>
- <details>
- <summary><strong>Agent (<code>apps/agent</code>)</strong></summary>
- <br/>
- | Variable | Default | Purpose |
- |---|---|---|
- | `HOST` / `PORT` | `127.0.0.1` / `4000` | Listen address. |
- | `AGENT_DATA_DIR` | `./data` | SQLite checkpoints and other runtime state. |
- | `AGENT_SKILLS_DIR` | bundled `skills/` | Directory of skills to mount. |
- | `AGENT_MAX_MODEL_CALLS` / `AGENT_MAX_TOOL_CALLS` | `40` / `60` | Budget per turn, shared with sub-agents. |
- | `AGENT_IDLE_TIMEOUT_MS` | `120000` | Longest silence from the model before a turn fails. |
- | `AGENT_RUN_TIMEOUT_MS` | `900000` | Upper bound for one turn. |
- | `AGENT_SESSION_TTL_HOURS` | `168` | Idle sessions are deleted after this. |
- | `AGENT_ATTACHMENT_TTL_HOURS` | `24` | Uploaded conversation files expire after this many hours. |
- | `AGENT_ATTACHMENT_SECRET` | random at startup | Optional stable HMAC secret, at least 32 characters; read URLs last up to 15 minutes. |
- | `AGENT_MAX_SEARCHES` | `5` | Search calls per turn, within the shared tool budget. |
- | `SEARCH_PROVIDER` | — | Optional `tavily`, `brave` or `searxng`; incomplete settings disable search and research skills. |
- | `SEARCH_API_KEY` | — | Required for Tavily and Brave. |
- | `SEARCH_BASE_URL` | provider default | Optional API base override; required for SearXNG, with no default public instance. Private hosts follow the model opt-in rule. |
- | `AGENT_BODY_LIMIT_BYTES` | `2097152` | Largest accepted JSON request. |
- | `AGENT_RATE_LIMIT_PER_MINUTE` | `300` | Requests per client address per minute; `0` disables the limit. |
- | `AGENT_ALLOW_PRIVATE_MODEL_HOSTS` | `false` | Allow model servers on private networks. Ignored in production. |
- See [`apps/agent/README.md`](./apps/agent/README.md) for the full list, how to write a skill, and a smoke test against a real model.
- </details>
- ## Open source and magic-resume.cn
- The editor and templates are the same; the hosted product runs the AI for you and adds accounts, sync and interviews.
- | | Open source (this repository) | magic-resume.cn |
- |---|---|---|
- | License | MIT | Free tier and paid plans |
- | Account | Not required | Required |
- | Resume storage | Your browser (IndexedDB) | Encrypted cloud storage, exportable at any time |
- | Multi-device sync | — | Yes |
- | AI | Polaris with your own model key, on the bundled agent | Managed models, mock interviews and memory |
- | Templates | 20 | 20 |
- ## Repository layout
- A pnpm and Turborepo monorepo. The shared schema, templates, design system, icons, utilities and the MCP server are published from [magic-resume-kit](https://github.com/Magic-Resume/magic-resume-kit) as `@magic-resume-kit/*` and pinned here by exact version.
- | Path | Description |
- |---|---|
- | `apps/web` | The editor and Polaris (Next.js) |
- | `apps/agent` | The agent behind Polaris (Fastify, DeepAgents, SQLite) |
- | `apps/agent/skills` | Skills and the agent's standing instructions, as Markdown |
- | `apps/landing` | Marketing site for magic-resume.cn (Astro) |
- | `apps/docs` | Documentation site (Mintlify) |
- | `packages/genui` | Polaris cards and widgets |
- | `packages/tsconfig` | Shared TypeScript configuration |
- ```bash
- pnpm dev # editor and agent
- pnpm build # build every workspace
- pnpm lint # lint and architecture checks
- pnpm test # tests
- pnpm --filter @magic-resume/agent test # one workspace
- ```
- ## Tech stack
- | Area | Technology |
- |---|---|
- | Editor | [Next.js 16](https://nextjs.org/), [React 19](https://react.dev/), [TypeScript](https://www.typescriptlang.org/), [Tailwind CSS 4](https://tailwindcss.com/), [Radix UI](https://www.radix-ui.com/), [Tiptap](https://tiptap.dev/), [Zustand](https://zustand-demo.pmnd.rs/), [i18next](https://www.i18next.com/) |
- | Documents | [React-PDF](https://react-pdf.org/), [PDF.js](https://mozilla.github.io/pdf.js/), [Zod](https://zod.dev/) schemas |
- | Agent | [DeepAgents](https://github.com/langchain-ai/deepagentsjs), [LangGraph](https://langchain-ai.github.io/langgraphjs/), [LangChain](https://js.langchain.com/), [Fastify](https://fastify.dev/), SQLite |
- | Tooling | [pnpm](https://pnpm.io/), [Turborepo](https://turbo.build/), ESLint, Docker |
- ## MCP
- [`@magic-resume-kit/mcp`](./docs/reference/mcp.md) is a Model Context Protocol server that lets AI coding tools such as Claude Code, Codex and Cursor read and edit resumes through schema-aware JSON Patch. It connects to a magic-resume.cn account; the local editor in this repository has no account API for it to reach.
- ```bash
- npx -y @magic-resume-kit/mcp@1.0.1 --help
- ```
- ## Roadmap
- - [x] Local-first editor with 20 templates
- - [x] Polaris on your own model key, with a self-hosted agent
- - [x] Resumable sessions, structured forms and approvals
- - [x] Import from PDF, Word, images and Markdown
- - [x] Attachments in conversations
- - [x] Optional web search with your own search API key, for company and role research
- Ideas and requests are welcome in [issues](https://github.com/Magic-Resume/Magic-Resume/issues).
- ## Contributing
- Contributions of every size are welcome: bug fixes, templates, skills, translations, docs and features. Read the [contributing guide](./CONTRIBUTING.md) for local setup, the branch workflow and the commit convention (gitmoji with Conventional Commits). For anything non-trivial, open an issue first to agree on the approach.
- <a href="https://github.com/Magic-Resume/Magic-Resume/graphs/contributors">
- <img src="https://contrib.rocks/image?repo=Magic-Resume/Magic-Resume" alt="Contributors" />
- </a>
- ## Star history
- <a href="https://star-history.com/#Magic-Resume/Magic-Resume&Date">
- <img src="https://api.star-history.com/svg?repos=Magic-Resume/Magic-Resume&type=Date" alt="Star history chart" width="600" />
- </a>
- ## License
- [MIT](./LICENSE) © 2026 [Kenmet Team](https://github.com/Magic-Resume)
- <!-- LINK GROUP -->
- [github-contributors-shield]: https://img.shields.io/github/contributors/Magic-Resume/Magic-Resume?color=c4f042&labelColor=black&style=flat-square
- [github-contributors-link]: https://github.com/Magic-Resume/Magic-Resume/graphs/contributors
- [github-forks-shield]: https://img.shields.io/github/forks/Magic-Resume/Magic-Resume?color=8ae8ff&labelColor=black&style=flat-square
- [github-forks-link]: https://github.com/Magic-Resume/Magic-Resume/network/members
- [github-stars-shield]: https://img.shields.io/github/stars/Magic-Resume/Magic-Resume?color=ffcb47&labelColor=black&style=flat-square
- [github-stars-link]: https://github.com/Magic-Resume/Magic-Resume/stargazers
- [github-issues-shield]: https://img.shields.io/github/issues/Magic-Resume/Magic-Resume?color=ff80eb&labelColor=black&style=flat-square
- [github-issues-link]: https://github.com/Magic-Resume/Magic-Resume/issues
- [github-license-shield]: https://img.shields.io/badge/license-MIT-white?labelColor=black&style=flat-square
- [github-license-link]: https://github.com/Magic-Resume/Magic-Resume/blob/master/LICENSE
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.