Artifacts: Live, Runnable Outputs
Artifacts are outputs Claude renders in a side panel next to the chat — a document, a chart, a working app, a diagram — that you can see, use, and iterate on, separate from the conversation text.
What you can make
- Mini web apps & tools — a calculator, a quiz, a form, a small interactive demo.
- Documents — structured write-ups you can refine and export.
- Visuals — charts, diagrams, and simple data dashboards.
- Code you can read and run.
Why it's powerful for non-developers
You can build something usable — "make me a tip calculator for a group dinner," "a dashboard from this CSV" — by describing it, then refine it conversationally ("add a service-charge field," "make the buttons bigger"). It's the clearest example of building with AI without writing code yourself.
How to work with Artifacts
- Ask for the thing, with specifics (purpose, inputs, look).
- Iterate in plain language — Claude updates the same artifact.
- Use it in the panel; export/share where supported.
Tips
- Be concrete about inputs/outputs and audience — same as good prompting.
- Iterate small. One change at a time is easier to get right.
- Verify any logic/numbers an artifact computes for important uses (Hallucinations).