AI agents
Files for agents
Everything on AI Canvas is also published as plain files, so an agent can read the catalog without a browser.
The files
| File | What is in it |
|---|---|
| /llms.txt | A short index: what AI Canvas is, how to install, every component and system with one line each. |
| /llms-full.txt | The long version, for an agent that wants everything in one read. |
| /r/registry.json | The shadcn registry index: every item, its files and its packages. |
/r/<name>.json | One registry item, the file the shadcn CLI installs. |
| /r/aicanvas-props.json | The documented props of every component in one file. |
llms.txt
llms.txt follows the llms.txt convention: a Markdown index an AI model can read in one go. Ours opens with what AI Canvas is and how to install, then lists the design systems with their install commands, the collections and every component. Point an agent at it when it needs to know what exists.
The registry
/r/registry.json is a standard shadcn registryindex, so any tool that understands shadcn registries understands it. Each item's own file at /r/<name>.json carries its source. Source needs a free account; see components.json.
The props file
/r/aicanvas-props.jsonholds every documented prop: name, type, whether it is required, its default and what it does. It is built from each component's own source comments, so it matches the code, and it needs no account. The MCP server serves the same data through get_component_props and checks files against it with validate_usage.