Playground
Try It: Simulated Claude Clients
Type the same request into each client below and compare what comes back. This is a fully simulated sandbox — nothing here calls the real Claude API. Every response is scripted and pattern-matched locally in your browser purely to illustrate how the three surfaces actually differ.
How to use this page
- Pick a tab: Claude.ai chat, Claude Desktop, or Claude Code.
- Type a request (or click one of the suggested prompts) and press Send.
- Try the same request in all three tabs to see how the response shape changes.
New to the three clients? Read the Claude Clients module first for the full explanation — this page is the hands-on companion to it.
A conversational answer in the browser. Substantial output (a doc, an app, a long piece of code) can render as a shareable Artifact.
Same chat experience as Claude.ai, but it can check local MCP connectors — local files, internal tools, and company systems — before answering.
Operates on a real project: plans the work, then reads, edits, and runs commands — shown here as tool calls and diffs instead of prose.
What to notice
- Claude.ai answers in prose and may offer an Artifact for anything substantial.
- Claude Desktop pauses to "check" a local file or MCP connector before answering.
- Claude Code always plans first, then shows tool calls and a diff — it's doing the work, not just describing it.