Claude Design, Slides and Docs in Any Chat: How to Use Them

Last updated: October 17, 2026

Since September 16, 2026, any Claude conversation - on the web, the desktop app, the iOS and Android apps, or in Claude Code - can produce a Claude Design canvas, a Claude Slides deck or a Claude Docs living document. Each one is saved to the Artifacts tab, where you or Claude can keep editing it. For developers, the useful parts are /design-sync plus /design for UI mockups grounded in your real components, /slides for turning a session into a review deck, and the Claude Docs connector for specs your teammates can edit. Watch the limits before you commit a workflow to it: Docs cannot be shared outside a Team or Enterprise organization, Design has no version history yet, and the standalone claude.ai/design app closes on December 14, 2026.

This guide covers what shipped, how to use each template from chat and from Claude Code, the settings that control it, the gotchas, and how it compares with ChatGPT's writing blocks and code blocks. Everything here comes from Anthropic's release notes, help center and Claude Code docs; the commands are quoted from the official docs rather than from a hands-on session.

What shipped on September 16#

The Claude release notes entry for September 16, 2026 is titled "Create designs, decks, and docs in any conversation". You can ask for a design, deck or document in any conversation, including in Claude Code and the Artifacts tab, and edit it with Claude or directly. Claude Design now works inside conversations with its full feature set, including on-canvas editing and design-system import. Claude Slides gives presentations their own starting point. Claude Docs is new: a way to write living documents with Claude and your team.

The structural change is that "artifact" is now the single container for everything Claude makes that you would put in front of someone. According to What are artifacts and how do I use them?, an artifact can start from one of five templates:

• Docs - living documents written with Claude and your team in real time.
• Slides - presentations built from notes, reports or the work already in the chat.
• Design - mockups, prototypes, one-pagers and landing pages built with your design system.
• Dashboards (beta) - charts built from data platforms or connected apps, each showing its query and last refresh.
• Motion (beta) - animated explainers that export as MP4.

Chat artifacts made before September 16 are now called legacy artifacts. They keep working and you can still publish and share them, but you cannot make new ones.

The same day, Anthropic also merged Cowork and chat into one Claude, and both changes share the Artifacts tab in the Claude sidebar. Our Cowork guide covers that half of the change.

Which plan gets what#

The release note says artifacts, including Design, Slides and Docs, are available on every plan including Free. The help center plan matrix breaks it down:

Two more plan boundaries matter for developers. Design systems are a beta feature on Pro, Max, Team and Enterprise, so Free users get the Design template but not saved design systems. And artifacts in Claude Code need a plan that includes Claude Code: the Claude Code artifacts docs list Pro, Max, Team and Enterprise.

Enterprise defaults are inconsistent across Anthropic's own pages. The release note said Design, Slides and Docs were in beta on Enterprise and off by default until an owner turned them on. The current artifacts article and the admin guide say they turned on by default on October 15, 2026. The migration article still lists Design and Slides as "On for Team plans, off for Enterprise plans", and the Claude Code docs say an Enterprise Owner turns each template on. If you are on Enterprise, do not rely on a default: check Organization settings > Artifacts.

%% title: Every entry point lands in the same artifact
flowchart LR
  A[Chat prompt] --> T{Template}
  B[Output menu] --> T
  C[Artifacts tab] --> T
  D["Claude Code /slides or /design"] --> T
  E[Claude Docs connector] --> T
  T --> S[Slides]
  T --> G[Design]
  T --> K[Docs]
  S --> R[Saved in Artifacts tab]
  G --> R
  K --> R
  R --> X[Edit, share, export]

Before you start#

In claude.ai (web, desktop, mobile):

1. Turn on Cloud code execution and file creation. On Free, Pro and Max it lives in Settings > Capabilities. On Team and Enterprise an owner sets it in Organization settings > Capabilities. Artifacts do not work without it.
2. There are three ways to start from a template: ask for one in any chat, select Output in the message box and choose a template, or pick one in the Artifacts tab. Everything you make lands in the Artifacts tab.
3. On iOS and Android you can ask for a design, deck or doc, view it in the Artifacts tab and edit it. Sharing settings can only be changed on web or desktop.

In Claude Code:

1. A Pro, Max, Team or Enterprise plan.
2. A session signed in with /login. Sessions that use an API key, a gateway token or a cloud-provider credential cannot publish artifacts.
3. The Anthropic API as the model provider. Artifacts are not available on Amazon Bedrock, Google Cloud's Agent Platform (Vertex) or Microsoft Foundry.
4. No CMEK, HIPAA or Zero Data Retention on the organization.
5. Claude Code v2.1.265 or later for /slides and /design.

Artifacts are off by default in the Agent SDK, the GitHub Action, MCP-server contexts, -p headless runs, and whenever CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC is set. If you script Claude Code in CI, expect it to write a local HTML file or tell you it cannot publish, rather than produce a deck.

How to make a deck with Claude Slides#

From chat. Ask in plain language, for example "turn this incident review into a six-slide deck for Thursday", or choose Output > Slides first. Claude builds the deck from your notes, attached reports or what is already in the conversation.

From Claude Code. Run /slides with a brief that says what the deck covers and who it is for. This is the example from the Claude Code docs:

/slides a quarterly review of the platform team's reliability work, for the engineering all-hands

Claude creates a Claude Slides artifact and gives you the link. If you run /slides with no brief, Claude asks what the deck should be about before it creates anything. You can also skip the command and ask in prose ("turn the migration notes into a deck for Thursday's review"), and Claude picks the matching template. If your organization has the Slides template turned off, /slides does not appear at all.

Edit and present. Open the deck in a desktop browser to edit or present it. You can edit any slide directly, ask Claude for changes, and present without leaving Claude. The editor has documented single-key shortcuts when you are not typing: T adds a text box, R a rectangle, O an oval, L a line, and F6 moves between editor areas (toolbar, slide, thumbnails, properties). While presenting, F toggles full screen and the number keys 1 to 9 jump to a slide. One trap: Delete or Backspace on a thumbnail deletes that slide immediately, though Cmd+Z (Ctrl+Z) brings it back.

Export. Decks export to PowerPoint and PDF, or go straight to Google Slides as editable files.

Where this earns its keep for developers: a long debugging or migration session already holds the timeline, the diffs and the decisions. /slides at the end turns that context into a review deck without re-explaining it, and the PPTX export means the deck can still go through whatever tool your org presents from.

How to make a design with Claude Design, and keep it on brand#

Get started with Claude Design describes Design as a conversation paired with a canvas. Here is the developer path that keeps output tied to your real components.

1. Pull your design system in with /design-sync. In Claude Code, run /design-sync from your repo. Claude reads your tokens and React components directly and builds a design system from them. You can also create a design system from a chat by uploading Figma files, decks, logos and fonts, and manage them in Settings > Design systems. The help center is explicit that import quality depends on the source: "A messy codebase or an incomplete file will show up in the output." For very large repositories, Anthropic recommends linking them from Claude Code with /design-sync rather than uploading through the browser.

2. Draft the design. In Claude Code, run /design with a brief that names what you want drawn:

/design a settings screen for a mobile banking app

Claude drafts artboards on one canvas and publishes it as a Claude Design artifact. In chat, either ask ("mock up the onboarding flow we just discussed") or choose Output > Design. From the Artifacts tab, pick a Design template.

3. Refine with the right tool. The help center gives a clear rule: use chat for structural changes and new sections, inline comments for one component ("make this button padding larger"), and direct canvas edits for quick visual nudges such as drag, resize and align. If an inline comment does not seem to get picked up, paste the same feedback into the chat - that is the documented workaround for an intermittent issue.

4. Hand off or export. Designs export as a .zip, PDF, PPTX, Google Slides or standalone HTML, or can be sent to tools such as Canva, Lovable, Miro, Netlify, Replit, v0 and Vercel. The handoff that matters most for developers is Handoff to Claude Code, which sends the design either to a local coding agent or to Claude Code on the web, so the implementation continues from the design instead of from a screenshot. From Claude Code, /design can also import a design into your codebase or export your code as a live prototype.

%% title: Design system round trip with Claude Code
sequenceDiagram
  participant Dev as Developer
  participant CC as Claude Code
  participant DS as Design system artifact
  participant CD as Claude Design canvas
  Dev->>CC: /design-sync
  CC->>DS: Read tokens and components
  Dev->>CC: /design settings screen
  CC->>CD: Draft artboards using DS
  Dev->>CD: Edit on canvas or comment
  CD->>CC: Handoff to Claude Code
  CC->>Dev: Implement in the repo

If you were already keeping agents on brand with a hand-written design contract, as in our DESIGN.md for AI agents guide, /design-sync is the productized version of that idea. The plain-file approach still has a role: the Claude Code docs say that when Claude builds an ordinary (non-template) artifact, its built-in design skill looks for an existing design system in your project, such as a ## Design system section in CLAUDE.md or a theme file, and treats it as higher precedence than its own choices. Your prompt still outranks both.

Migrating from standalone claude.ai/design#

The standalone app keeps working until December 14, 2026, after which claude.ai/design redirects to Claude. The migration article lays out the steps:

1. Open the Artifacts page in Claude.
2. Click Migrate team design systems in the "Claude Design lives here now" banner (or choose it from the Claude Design menu at the top right if you closed the banner).
3. Click Migrate to confirm. Individual plans skip this confirmation.
4. Check the Migrated and Couldn't be migrated lists. A "Fix in Claude Design Standalone" link runs a check that names the files in the way, such as files that are too large or have unusable names.

One migration brings over every design system in the organization, including other people's, and anyone who can use Claude Design can start it. Published design systems that are not private become visible to everyone in the organization. What does not come across:

• Empty design systems, and systems made from a built-in starter theme, even if edited.
• Your chats and comments from the standalone app. They are deleted after it closes, so save anything you need now.
• Public links to standalone projects, which stop working.

Projects themselves stay in the standalone app until it closes; the Artifacts page lists up to 60 of them, and they open in the standalone version. Anthropic says more details on moving projects are coming. Zero-data-retention organizations cannot turn on artifacts, so they cannot migrate at all; they should download each project (Share > Project HTML > Project archive > Export) to get a free .zip of every file. For the background on what standalone Design could do when it launched in April, see our Claude Design developer guide.

How to write a living doc with Claude Docs#

Get started with Claude Docs describes a doc as a rich-text document saved to your Claude account, with headings, tables and multiple tabs. Claude can draw on your files, memory, projects, skills and connected apps.

From chat. Start your request with /docs, choose Output > Docs, or simply ask: "Turn the plan we just worked through into a product spec I can share with the team." Claude asks a few clarifying questions first, drafts on screen, and leaves comments explaining its choices or asking you something.

From the Artifacts tab. Pick a Docs template from the gallery and describe the doc.

Editing together. Nothing is locked while Claude works. You type directly, ask Claude in the conversation, or select text and mention @Claude in a comment. Claude replies in the thread, makes the change and explains it. People with edit access see each other's edits in real time.

%% title: How an @Claude comment edits a doc
sequenceDiagram
  participant U as Editor
  participant Doc as Claude Docs
  participant C as Claude
  U->>Doc: Select text, comment @Claude
  Doc->>C: Request with editor permissions
  C->>Doc: Apply edit, attributed to Claude
  C->>Doc: Reply in thread explaining change
  Note over U,C: A viewer cannot trigger edits

The permission model is the part worth understanding. Claude works on a doc only when a signed-in person asks, and it always acts with that person's permissions, so it cannot edit a doc on behalf of someone with view-only access. Every change is attributed to whoever made it, person or Claude.

From Claude Code. Docs is not a slash command there. It arrives as a claude.ai connector, and /mcp lists it as claude.ai Claude Docs. Once it is connected, a request for a document meant for other people - a spec, a proposal, a write-up of the plan you just worked through - goes to Claude Docs, and Claude hands you a link (on desktop the doc opens in the side panel). A document that belongs in the codebase, like a README, stays a file. If you want a file for something Claude would otherwise put in Docs, name the format: .docx, or a Markdown file in the repository.

Export. Docs export to Word, PDF, Markdown and Google Docs. You can also ask Claude to turn a doc into a Claude Slides deck.

Configure it in Claude Code#

Most of the control surface is in Claude Code settings. Both JSON snippets below parse cleanly, and every key comes from the Claude Code docs.

Keep repo docs as files. If you want a project to never route documents to Claude Docs, deny the connector by name. deniedMcpServers works in any settings scope, and denylists from every scope merge:

{
  "deniedMcpServers": [
    { "serverName": "claude.ai Claude Docs" }
  ]
}

You can also flip the connector off interactively with the /mcp toggle. To drop every claude.ai connector that Claude Code fetches itself, the MCP docs document "disableClaudeAiConnectors": true instead.

Turn artifacts off entirely. Any one of these works:

• "enableArtifact": false in a settings file (the /config Artifacts row writes it to your user settings). Honoring it in a project's .claude/settings.json or .claude/settings.local.json needs v2.1.242 or later.
• CLAUDE_CODE_DISABLE_ARTIFACT=1 in the environment.
• Artifact in permissions.deny.

{
  "enableArtifact": false
}

Once artifacts are off via --settings, the environment variable or managed settings, no settings file turns them back on, and an "enableArtifact": true in a project file never re-enables them. For pushing these policies to a whole team, our Claude Code plugins for teams guide covers managed settings rollout.

Day-to-day controls:

Gotchas and limits#

Claude Docs

• No comment-only access level. People are viewers (read) or editors (read, edit, comment, export).
• No email invitations on any plan. On Team and Enterprise, no sharing outside the organization by link either. Free, Pro and Max can share by link.
• Opening a shared doc requires a Claude account.
• Only the owner can rename or delete. Deletion is permanent with no trash, and the doc disappears for everyone.
• Charts and diagrams do not auto-refresh, even from connected apps. Ask Claude to pull the latest data.
• Edits and comments inside a doc are not yet recorded in the Compliance API.

Claude Design

• No version history yet. The workaround in the help center is to ask Claude to "save what we have and try a completely different approach."
• Multi-person editing of one design is still basic and may not work reliably.
• Inline comments can occasionally fail to show or be missed. Paste the feedback into chat.
• Placing new layers, artboards or shapes needs a mouse or trackpad; ask Claude to add them if you work keyboard-only.
• Design import is only as good as its source.

Everywhere

• Designs, decks and docs count toward your existing usage limits, shared with the rest of Claude including Claude Code. Claude Design no longer has a separate weekly allowance. Usage-based Enterprise plans bill at standard API rates.
• Sharing settings cannot be changed in the mobile apps.
• Artifacts are not available through third-party cloud platforms (Bedrock, Vertex, Foundry).
• Artifact data storage is capped at 20 MB per artifact and is text only.
• CMEK conflict: the admin guide lists artifacts as available for CMEK organizations, while the Claude Code docs list CMEK as blocking artifact publishing from Claude Code. If your organization uses CMEK, test the Claude Code path before planning around it.

How it compares to ChatGPT writing blocks and code blocks#

The closest documented equivalent is ChatGPT's writing blocks and code blocks. They do the in-chat job well: a writing block is an editable draft with formatting where you can revise a selection or the whole draft, open a full-screen editor, undo AI edits, send supported email drafts, and save supported document drafts to your Library. Code blocks preview HTML, React, SVG, Mermaid and Vega charts and run Python in a sandbox. OpenAI notes that available actions vary by plan, device, workspace settings, model and rollout.

This table compares only what the two help articles document:

How to read it: ChatGPT's blocks are a strong fit for quick single-author drafts and code previews that live in the conversation. Claude's templates are separate, persistent objects with collaborators, a presentable deck editor and a design canvas wired into a coding agent. The cost of that extra surface is extra rules - plan gates, organization sharing limits and the Enterprise toggles above. For the plan-level picture, see ChatGPT Plus vs Claude Pro for developers.

FAQ#

Is Claude Slides available on the Free plan?#

Yes. Free includes creating artifacts in chat and starting from the Docs, Slides or Design templates. The Dashboards template needs Pro or higher, Motion needs Team or Enterprise, and saved design systems need Pro or higher. "Cloud code execution and file creation" must be turned on in Settings > Capabilities.

How do I make a deck or design from Claude Code?#

Run /slides <brief> or /design <brief> on Claude Code v2.1.265 or later, in a session signed in with /login on a Pro, Max, Team or Enterprise plan using the Anthropic API. Claude Docs arrives as the claude.ai Claude Docs connector rather than a command, so you just ask for a spec or proposal meant for other people.

What happens to my claude.ai/design projects?#

The standalone app closes on December 14, 2026. Migrate design systems now from the Artifacts page with "Migrate team design systems". Projects stay in the standalone app until it closes, chats and comments are not carried over, and you can download each project as a .zip project archive.

Can I share a Claude Doc with someone outside my company?#

Not on Team or Enterprise plans, by link or by email. Free, Pro and Max can share by link, and anyone opening the doc needs a Claude account. No plan supports email invitations for docs.

Does using Design, Slides or Docs cost extra?#

There is no separate allowance. Everything draws from the same usage limits as chat and Claude Code. Usage-based Enterprise plans bill at standard API rates.

Sources#

• Claude release notes - September 16, 2026
• What are artifacts and how do I use them?
• Get started with Claude Docs
• Get started with Claude Design
• Set up your design system in Claude Design
• Migrate from standalone Claude Design to Claude
• Artifacts admin guide for Team and Enterprise plans
• Claude Code docs: Artifacts
• Claude Code docs: MCP
• Claude Code docs: Managed MCP
• OpenAI Help: Working with writing blocks and code blocks in ChatGPT

Continue Reading#

• Claude Design developer guide - the April deep dive on standalone Claude Design, repo scans and coding-agent export
• Cowork: Claude Code for everyone - Cowork merging into one Claude, shipped the same day as these templates
• DESIGN.md for AI agents - keeping agents on brand before design systems were built in