Skip to content

llms.txt

ThemeKit ships a static context layer for large language models — plain-text files, following the llms.txt convention, that any AI editor or agent can fetch to generate correct, token-bound ThemeKit code without guessing APIs or hardcoding colors.

It’s the drop-in companion to the MCP server: the MCP gives your editor live, on-demand tools, while these files give any tool a complete snapshot it can attach or fetch in one shot.

Point your tool at the hosted URL, or download the file and attach it:

https://isamercan.github.io/ThemeKit/llms.txt
https://isamercan.github.io/ThemeKit/llms-full.txt
https://isamercan.github.io/ThemeKit/llms-components.txt
https://isamercan.github.io/ThemeKit/llms-patterns.txt
  • Claude · ChatGPT · Cursor · Copilot — attach llms.txt as project context, then ask for a screen. For deep API detail, add llms-components.txt.
  • URL-fetching agents — hand them the link above; llms.txt cross-links the rest.
  • The one rule that keeps generated code correct — read the theme from the environment and never hardcode a color:
@Environment(\.theme) private var theme // not Theme.shared inside a subview
theme.text(.textPrimary) // not .foregroundStyle(.blue)
Theme.SpacingKey.md.value // not .padding(16)
Theme.RadiusRole.box.value // not cornerRadius: 12

If your editor speaks MCP, the ThemeKit MCP server exposes the same data as on-demand tools (get_component_api, get_design_tokens, generate_theme, a11y_audit, Figma round-trip…) — focused context pulled exactly when the model needs it, instead of one large paste. The llms.* files and the MCP server are built from the same source, so they always agree.

Static context

One fetch or paste, works everywhere. Best for tools without MCP, or for a complete snapshot in a single file.

Live tools (MCP)

On-demand, focused API/token lookups while the model codes. Best for MCP-capable editors. Set it up →