Artifact Tools
Artifact tools enable your AI agent to create and edit rich documents during conversations. Unlike regular chat responses that scroll away, artifacts persist as editable documents that can be refined, versioned, and shared.
Create Artifact
The Create Artifact tool lets the AI generate substantial documents: reports, summaries, code files, creative content, and structured documents. When the AI determines that creating a document would be helpful, it generates an artifact with a descriptive name and content. The artifact appears in the chat interface where users can view, edit, and download it.
No special configuration is required—simply enable the tool on your agent and the AI will use it when appropriate.
Edit Artifact
The Edit Artifact tool allows the AI to modify existing artifacts based on user feedback. It can update content, fix errors, expand sections, and revise formatting. Each edit creates a new version while keeping previous versions accessible in the history.
When the AI edits an artifact, it identifies the existing document, makes changes while preserving overall structure, and creates a new version. Users can always browse back to see what changed.
Working with Artifacts
Every artifact maintains a complete version history. Each edit—whether from the AI or manual changes—creates a new version. You can browse previous versions, see what changed, and restore older versions if needed.
Users can edit artifacts directly using the built-in editor. Make corrections, add personal notes, or restructure content as needed. Saving your changes creates a new version, preserving your original in the history.
Artifacts can be exported in multiple formats. Word (.docx) works well for further editing in Microsoft Office or sharing with collaborators. PDF is ideal for final documents, printing, or distribution where editing shouldn't occur. Markdown (.md) suits technical documentation and version-controlled repositories.
Share artifacts with team members by clicking the Share button and selecting who should have access. Shared users can view the artifact and add comments, though only the owner can edit directly.
When an artifact is valuable beyond the current conversation, save it to a project by clicking Save to Project and selecting your target. The artifact becomes part of that project's file collection, available for reference in future conversations.
Best Practices
Artifacts work best for longer documents that users will reference later or need to revise. Reports, summaries, and structured content are good candidates. For simple, short answers or one-time information, regular chat responses are more appropriate.
The AI should give artifacts clear, descriptive names that make them easy to find later. Encourage this in your agent's system prompt if needed.
Don't worry about losing content through edits—versions are automatic. Users can experiment freely knowing they can always restore an earlier version.
Example Interactions
When a user asks for a summary report, the AI creates an artifact rather than just responding in chat:
User: Create a summary report of our Q3 sales data
AI: I'll create a detailed report for you. [Creates artifact: "Q3 Sales Summary Report"]
Later, the user can ask for modifications:
User: Can you add a section about regional breakdown?
AI: I'll update the report with regional data. [Edits artifact: adds new section]
Users can also edit artifacts directly—clicking into the document, making changes, and saving—then continue the conversation for additional AI edits.