by JimLiu
Open, agent- and model-agnostic artifacts, all saved locally as plain files — edit, comment on, and export slides, UI designs, design systems, whiteboards, sheets, dashboards, docs, and motion.
Available after the catalog security scan passes.
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM lists this project for discovery and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
# Add to your Claude Code skills
git clone https://github.com/JimLiu/open-artifactGuides for using ai agents skills like open-artifact.
See how open-artifact compares with popular alternatives.
English · 简体中文
Formerly
baoyu-design. This project started as baoyu-design (the old URL redirects here), which brought claude.ai-style Artifact types to local coding agents as one skill per type. Open Artifact rebuilds them as an open, agent-agnostic format: one local daemon, one page frame, one/open-artifactskill, and a plugin for each agent app.
Open Artifact lets any coding agent make things people can open, edit and comment on, not just files. You ask your agent for a slide deck, a UI design, a design system, a whiteboard, a spreadsheet, a data dashboard, a doc or a short animated film. The agent writes a plain folder of files. A local page opens it in a real editor, where you can fix a title, move a sticky note, leave comments, and export. Then "Send to agent" sends your comments back to the agent's session.
https://github.com/user-attachments/assets/9f34dc69-2e90-4118-92fd-63f36d7bc5f9
Everything runs on your machine. No account, no upload, no specific model or vendor.
you ──"/open-artifact slides a 6-slide pitch"──▶ agent (Claude Code, Codex, Cursor, Copilot, …)
│ writes files
▼
./my-deck/project/deck.json + slides/*.html
│ served by the local daemon
▼
you ◀── edit · comment · export (PDF/PPTX/PNG/HTML) ── http://127.0.0.1:<port>/a/<id>
└──────────────── "Send to agent" ──────────────▶ comments arrive back in the agent's session
| Type | What you get | Folder marker | Exports |
|---|---|---|---|
| Slides | A 1920×1080 deck: slide rail, overview, speaker notes, builds, transitions, presenter mode | project/deck.json |
PDF, PPTX, HTML, PNG |
| Design | An infinite canvas of live .dc.html artboards: screens, components, click-through prototypes, notes, pages |
project/canvas.json |
PNG, JPEG, PDF, HTML |
| Design System | A brand book: tokens, colors with contrast checks, type specimens, live component previews, assets. Designs and decks can install it | project/design-system.json |
— |
| Whiteboard | A shared infinite board: stickies, shapes, arrows, ink, images, tables, embeds, live cursors, timer | db/board/meta.json |
PNG |
| Sheets | A shared spreadsheet: several sheets of cells, a full formula engine, number formats, borders, merged cells, tables with filters, dropdowns, checkboxes, conditional formats, frozen and grouped rows, charts from cells, CSV import, cell history | sheet/sheet.json |
XLSX, CSV |
| Data | A live dashboard or report: datasets from CSV/JSON/text uploads, live MCP connector queries or calculations, drawn by a plain HTML/CSS/d3 page; Sources shows where each number comes from; filters, hand edits, remix drafts | db/dash/meta.json |
PDF, HTML, JSON |
| Docs | A living document in tabs: headings, lists, tables, pictures, dates, mentions and dropdowns; suggested edits to accept or reject, comments, a version history per tab; imports Markdown, Word and PDF | doc/doc.json |
Markdown, Word, PDF, HTML |
| Motion | A short animated film: scenes in HTML/CSS/JS drawn from time, a generated score, rough cut then final, captions and voiceover; retype, recolor, refont, recut and restyle by hand; comments pinned to a moment or a spot in the frame | db/film/meta.json |
MP4, GIF, HTML, ZIP |
Each type folder is a self-contained skill: SKILL.md tells the agent how to write it, reference/ documents the
file format, artifact-type.json describes it to hosts, and scripts/ (built) holds the CLI and editors.
Install Node.js 22.18+, then download open-artifact-plugins.zip from the Assets section of a
GitHub Release. Extract it to a permanent location and use the
extracted open-artifact-plugins/ folder in the steps below. The plugins include the skill, daemon and editors.
DeepSeek Harness installs its prebuilt package directly from npm.
| Agent app | Install |
|---|---|
| Claude Code (CLI, desktop, IDE) | Add the extracted root as a marketplace, then install open-artifact@open-artifact. |
| Codex (CLI, app) and ChatGPT desktop | Add the extracted root with codex plugin marketplace add, then install Open Artifact in Plugins. |
| Cursor | Copy the extracted cursor/ folder to ~/.cursor/plugins/local/open-artifact, then run Developer: Reload Window. |
| GitHub Copilot CLI | Add the extracted root as a marketplace, then install open-artifact@open-artifact. |
| GitHub Copilot in VS Code | Run Chat: Install Plugin from Source and select the extracted copilot/ folder. |
| Antigravity | Install the extracted antigravity/ folder with agy plugin install or copy it to ~/.gemini/config/plugins/open-artifact, then restart the app. |
| ChatGPT on the web | Run the prebuilt connector from the extracted chatgpt/ folder and connect it over HTTPS. |
| DeepSeek Harness | In Plugins, choose Add plugin and enter dsh-open-artifact as the npm package. In the CLI, run dsh plugin --profile web add dsh-open-artifact for the web profile. |
To upgrade, download the new ZIP, replace the extracted folder at the same path, and follow your app's update steps in the linked guide. If your marketplace still points to a source checkout, register the extracted folder using the migration steps in your app's guide.
In DeepSeek Harness, use Update on the Open Artifact page when a new version is offered. If you previously
linked a local checkout, remove the linked plugin in Plugins and add dsh-open-artifact from npm.
In a new session, try one of these prompts.
/open-artifact slides a 6-slide pitch for this repo
/open-artifact design three landing page options for a coffee shop
/open-artifact whiteboard a retro board for last sprint
/open-artifact sheets a monthly budget with categories and totals
/open-artifact data a dashboard of signups by month from signups.csv
/open-artifact docs a hiring plan for 2027 with roles, budget and timeline
/open-artifact motion a 20-second launch animation for this app
The plugin adds the Artifact, ArtifactData and ArtifactComments MCP tools, the skill, and hooks that bring
page comments back into the session. How each app gets comments back, and what was tested in each, is in
plugins/README.md.
In an app that shows MCP Apps, the agent also gets a
ShowArtifact tool: it puts the artifact's live page in the conversation, where you can edit and comment without
leaving the chat, go full screen, open it in a browser or send your comments to the agent. Apps that don't show
MCP Apps never see the tool.
Copy codex/skills/open-artifact/ from the extracted plugin bundle to your agent's skill directory. It carries
the daemon, and the agent drives it from the command line. From the extracted bundle root, you can try:
node codex/skills/open-artifact/scripts/open-artifact.mjs artifact '{"action":"list","scope":"types"}'
After building from source, every built type folder runs by itself, without the daemon:
node artifact-types/design/scripts/cli.mjs open ./my-design --type design # starts a local server, prints the URL
node artifact-types/design/scripts/cli.mjs wait ./my-design # blocks until someone presses "Send to agent"
Other commands: check, generate, upload, context, comments, history, share, status, stop, … Run the
CLI with no arguments for the full list.
From a source checkout, install dependencies and build the workspace and plugin folders.
npm ci
npm run build
node scripts/build-plugins.mjs
For local plugin development, use dist/plugins/ wherever the installation guides use the extracted folder.
npm run typecheck
npm test # unit tests (node --test)
npm run e2e # end-to-end: drives the built types with the CLI and Chrome, see e2e/README.md
The Rust crates in crates/ compile to WebAssembly. Their output is committed under packages/wasm-*/generated/,
so you only need Rust when you change a crate:
node scripts/build-wasm.mjs # needs cargo + wasm32-unknown-unknown, wasm-bindgen, wasm-opt
node scripts/build-npm.mjs stages the npm packages in dist/npm/: the @open-artifact/* libraries (compiled to
JavaScript with type declarations), @open-artifact/cli (the open-artifact command with the daemon vendored) and
dsh-open-artifact. To release:
node scripts/changelog.mjs for the commits since the last tag, and add a ## [x.y.z] - date entry at the top
of CHANGELOG.md that says what changed for people using it (Added / Changed / Fixed), plus its
compare link at the bottom.v0.9.0). The Publish to npm workflow requires a
changelog entry for the tag. Iopen-artifact is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by JimLiu. Open, agent- and model-agnostic artifacts, all saved locally as plain files — edit, comment on, and export slides, UI designs, design systems, whiteboards, sheets, dashboards, docs, and motion. It has 135 GitHub stars.
open-artifact's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
Clone the repository with "git clone https://github.com/JimLiu/open-artifact" and add it to your Claude Code skills directory (see the Installation section above).
open-artifact is primarily written in TypeScript. It is open-source under JimLiu on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh open-artifact against similar tools.
No comments yet. Be the first to share your thoughts!