A few days ago, the old Workbench in the Claude Console went dark. Now its successor is ready: Playground. Anthropic didn’t just rename the tool – it noticeably expanded it.
What Playground can do
The key difference: Playground supports every parameter of the Messages API. So whatever you could set in code, you can try here first – without writing a line.
On top of that come templates that demonstrate individual API features. Code execution and web search are two of them. You don’t start from zero; you start with an example and adjust it.
And the point that helps most when learning: for every run, Playground shows the full SDK request and the API response. So you don’t just see the result – you see exactly what the request would look like in code. That closes the gap between trying things out and building them.
You’ll find all of it at platform.claude.com/playground or straight through the Console.
The context
This fits what Anthropic had announced beforehand. On August 17, the old Workbench was sunset, along with the experimental prompt tool APIs. Saved prompts, variables, and evals from the old environment don’t carry over automatically – if you still had something there, you had to export it first.
If you missed the switch, Playground is the new home now. Just without the old saved bits.
My take
The name Playground is well chosen. A workbench is a place to work; a playground invites you to try things – and that’s exactly the point here. For me, having each request shown in plain text right next to it is the real win. You learn an API fastest when you see what your clicking around means in code. For people just starting out with the Claude API, that’s the shorter path from first attempt to first line of their own.
Sources: