Hello, devs! 😎
Have you ever used Artifacts in Claude?
An Artifact is a dedicated window right next to the chat where Claude renders "substantial" content (such as code, HTML, Markdown, SVG, Mermaid diagrams, React components, and more) instead of dumping everything right into the middle of the conversation.
If you use LLMs daily to program, document, or design software architectures, you know exactly where it hurts. It’s that feeling of getting lost in a mountain of text within the chat window. You ask for a script, and the model spits out 200 lines of code. You ask for a minor change, and here come another 200 lines, cluttering your screen and causing you to lose track of the conversation context.
To improve this experience, you can start using Claude's Artifacts. They split the screen into two parts: on the left, you keep the traditional chat to give commands, ask questions, and brainstorm ideas. On the right, a clean, isolated workspace opens up where the final file is created.
How to Enable It?
Go to Settings → Capabilities (or Feature Previews, depending on your plan) and ensure that Code execution and file creation (or Artifacts) is toggled on. It is available across all tiers (Free, Pro, Max, Team, Enterprise).
How Does Claude Decide to Isolate Content into an Artifact?
Claude automatically creates an Artifact when the content is long, self-contained, and reusable. For instance: a script with more than 20 lines, a complete document, an HTML page, or a diagram. Short or purely conversational content (like an explanation or a 5-line snippet) stays in the regular chat. If you want to force its creation, simply ask explicitly: "Put this in an artifact."
Each Artifact operates in isolation:
HTML/JS/CSS code runs in its own sandbox inside the window.
Markdown documents are treated as beautifully rendered text files.
Mermaid diagrams are compiled visually on the spot.
This means the Artifact's content never "pollutes" the rest of your conversation, and you can easily copy, download, or publish it separately.
The main supported types are: code, documents (Markdown/text), interactive HTML pages, SVG, React components, Mermaid diagrams, and even freshly generated PDFs and Office files (.docx, .pptx, .xlsx).
Iterating with Artifacts
You don't need to rewrite everything for every single adjustment!
How Does It Work in Practice?
Once the Artifact is created, you simply keep typing in the regular chat: "change the button color to blue," "add a column for the total," or "summarize the second paragraph." Claude updates the existing Artifact, and the window immediately reflects the latest changes.
Version Control: Every update generates a new version, accessible via a version selector (usually at the bottom of the Artifact window). You can jump back to previous versions without losing any work.
Inline Editing: For Markdown documents specifically, you can use targeted editing. Select the piece of text you want to change, click "Edit with Claude," and describe your modification. Claude will edit only that specific snippet, saving you from explaining the context all over again in the chat.
Branching Conversations: Editing an earlier message in the chat creates a new branch—a brand-new set of Artifacts associated with that version of the conversation, allowing you to explore alternative technical paths without destroying your previous work.
Pro Tip: Claude tends to be much more efficient (and precise) when you ask for incremental refinements instead of requesting to rewrite the entire Artifact from scratch every time. This preserves context history and reduces the risk of content truncation.
Visualization with Mermaid
Mermaid is a JavaScript-based diagramming and charting tool that uses Markdown-inspired text definitions. Claude renders these visually right inside the Artifact window.
How to Ask For It:
"Create a flowchart showing the order approval process, from the customer to payment."
"Draw a sequence diagram of the OAuth authentication between the client, auth server, and API."
"Turn this explanation into a mind map using Mermaid."
Claude writes the Mermaid code (text syntax), and the Artifact automatically compiles it into a visual diagram—meaning you don't even need to know the syntax yourself.
sequenceDiagram
autonumber
participant U as User
participant C as Client (App)
participant AS as Authorization Server
participant API as API (Protected Resource)
U->>C: Initiates login
C->>AS: Redirects to /authorize<br/>(client_id, redirect_uri, scope, state)
AS->>U: Displays login and consent screen
U->>AS: Enters credentials and grants access
AS->>C: Redirects with authorization_code<br/>(code, state)
Note over C,AS: Code exchange for token
C->>AS: POST /token<br/>(code, client_id, client_secret, redirect_uri)
AS->>AS: Validates code and client credentials
AS->>C: Returns access_token<br/>(+ refresh_token, expires_in)
Note over C,API: Client accesses protected resource
C->>API: GET /resource<br/>Authorization: Bearer access_token
API->>AS: Validates access_token<br/>(introspection or local verification via JWKS)
AS->>API: Token is valid (scopes, expiration)
API->>C: Returns protected data
Note over C,AS: When the token expires
C->>AS: POST /token<br/>(grant_type=refresh_token, refresh_token)
AS->>C: Returns new access_token
Once generated, you can iterate on it seamlessly just like code: "add a validation step between X and Y," "reverse the arrow directions," or "switch this to a sequence diagram instead of a flowchart."
Diagram Types that Mermaid (and Claude) Supports Well:
- Flowcharts (flowchart or graph)
- Sequence Diagrams (sequenceDiagram)
- Class and Entity-Relationship Diagrams (classDiagram, erDiagram)
- Gantt Charts (gantt)
- Mind Maps (mindmap)
- State Diagrams (stateDiagram)
If this article helped streamline your workflow or shed light on new ways to use Claude daily, share it with your dev friends or your engineering team who might still be wrestling with cluttered, generic chats!
Liked it? You can find me on LinkedIn and GitHub:
gabideutner
Thanks for reading!
Top comments (0)