Here's a great use case for Claude and the Basecamp CLI. Pull together some inspirational photos of iconic furniture from around the world, and upload them to the appropriately named galleries inside a Basecamp project. Here's a quick walkthrough:

the low stakes are what make this demo work, a chair going in the wrong gallery costs nothing. the real test is pointing the same cli at classification where a wrong bucket actually costs something, like sorting tasks or client messages into the wrong project.

The interesting part isn't the upload, it's letting Claude decide which country's chairs count as iconic without a style guide.

This is such a smart workflow—turning inspiration into organized, reusable project assets is a game changer

The interesting part is the agent actually writing into a live project. Finding photos is easy to demo. I'd spend more time on what it's allowed to create, where it can upload, and how to undo a bad write.

The reason CLIs turned out to be the best agent surface: every command is already a documented, composable unit with arguments and a return code. An MCP server is that plus a schema. Most products already had a good agent interface, they just called it a CLI and never said so.

finally, mood boards with shell access

Using the CLI to automate asset organization is a smart way to actually clear the desk, though I suspect the AI is going to spend more time arguing about what counts as iconic mid-century design than actually sorting the files