@pantoken/ai

Agent assets for projects that use pantoken (not for developing pantoken itself). It ships an AGENTS.md, an llms.txt, editor and agent rule files (Cursor, Copilot, Windsurf), and a Claude Code bootstrap skill — plus an installer CLI that drops them into a consumer repo at the conventional paths.

Install

npm i -D @pantoken/ai

Usage

Run the installer to write the assets into a repo:

# Everything (AGENTS.md, llms.txt, Cursor, Copilot, Windsurf, Claude skill):
npx pantoken-ai init

# One tool:
npx pantoken-ai init --tool cursor
npx pantoken-ai init --tool claude --dir ./my-app

Tools: agents, llms, cursor, copilot, windsurf, claude, or all.

Or install programmatically:

import { installAgentAssets, AGENTS_MD, LLMS_TXT } from "@pantoken/ai";

installAgentAssets("cursor", "./my-app"); // → .cursor/rules/pantoken.mdc

The assets teach an agent how to consume pantoken: the token model (--instui-*), which package to reach for per framework and platform, the pantoken generate <target> CLI, and the conventions (prefer var(--instui-*) references, never invent token names).

API

• installAgentAssets(tool, dir?): string[] — write pantoken's agent assets for a tool into a consumer repo. tool is a specific AgentTool or "all"; dir defaults to ".". Returns the paths written.
• AGENT_TOOLS: readonly AgentTool[] — every installable tool key.
• AgentTool — the tool union: "agents", "llms", "cursor", "copilot", "windsurf", and "claude".
• AGENTS_MD, LLMS_TXT — the AGENTS.md and llms.txt contents as strings.
• ASSETS — the raw asset map (re-exported from the generated assets module).

License

MIT