This is the AI and agents part of the Atomus documentation: rules for agents, the skill, the Atomus MCP server, lint rules, the shadcn registry, evals, Figma MCP and llms.txt. --- # Atomus is built for humans and agents > What Atomus gives AI agents today (guidelines, skill, MCP server, lint rules, shadcn registry, Storybook, Agent kit, llms.txt, Figma MCP rules), how it is measured, and what is coming next. URL: https://docs.atomus.io/ai/ Most design systems are written for people and then scraped by AI tools. Atomus is written for both. Every rule a designer or developer reads on this site is also published in a form agents can follow literally. The props, tokens and rules agents use are **generated from the same sources** as the Figma file and the React package, so they can’t drift. The result: Claude Code, Cursor, Copilot, Figma Make and the Figma MCP server pick the right component, use only real props, style with semantic tokens and keep one primary action per view. You review the result instead of fixing it. ## Available now Guidelines for agents Plain-Markdown guidelines with an entry point (`Guidelines.md`), a component catalogue with decision trees, imperative **Do / Forbidden** rules, and a **React API** section on every component generated from the TypeScript source. Atomus skill An [Agent Skill](https://agentskills.io/) with the workflow, references (components, tokens by intent, page patterns, Figma) and a validator that flags raw colours, primitive tokens and invented props. `npx skills add StanVisionAgency/atomus`. AGENTS.md template Paste-in rules for product repos that use Atomus, read by Claude Code, Cursor, Copilot and Codex. llms.txt and Markdown pages `/llms.txt`, `/llms-full.txt`, `/llms-small.txt`, topic subsets, and every page as Markdown at `…/index.md`. Figma MCP rules How an agent reads Atomus designs through the Figma MCP server: modes on the top frame, Code Connect first, variables to tokens, never paste hex. Code Connect mappings All 21 React components are mapped to their Figma component sets, ready to publish to Dev Mode. Atomus MCP server Components, props, tokens by intent, page patterns, Figma to code, validation, setup files and brand ramps as MCP tools. [Set it up](https://docs.atomus.io/ai/mcp/). ESLint and Stylelint rules The Atomus rules as lint rules for editors, CI and agent loops, with SARIF output. [Lint rules](https://docs.atomus.io/ai/lint/). shadcn registry Every component, four app screens, an AI chat block and 30 website sections as source, for the shadcn CLI, the shadcn MCP server and v0. [Registry](https://docs.atomus.io/ai/registry/). Storybook Stories for every React component with theme, brand and radius switches, a component manifest and an MCP server for agents. Agent kit 12 components for AI products — prompt input, messages, streaming text, reasoning, tool calls, approvals, citations, suggestions, model selector, feedback, AI label and context meter — on the same tokens and modes. See the [Conversation](https://docs.atomus.io/components/ai-conversation/) demo. ## Coming next | What | Status | | --------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **npm packages** `@stanvision/atomus-react` and `@stanvision/atomus-tokens` | Publishing soon. Until then, install from the [repo](https://github.com/StanVisionAgency/atomus). | | **Figma Make kit** | After the npm packages publish. See [Figma Make kit](https://docs.atomus.io/ai/figma-make/). | | **Evals results** | The [evals harness](https://docs.atomus.io/ai/evals/) scores agent output against the rules (real imports and props, lint, axe, no raw hex). The first scorecard per agent is coming with the first run. | ## Start here [Claude Code, Cursor and Copilot](https://docs.atomus.io/ai/coding-agents/) Install the skill, add the AGENTS.md template, validate the output. [Atomus MCP server](https://docs.atomus.io/ai/mcp/) Components, tokens, patterns and validation as tools, local or remote. [Figma MCP and Code Connect](https://docs.atomus.io/ai/figma-mcp/) Turn Atomus designs into Atomus code, not div soup. [llms.txt](https://docs.atomus.io/ai/llms-txt/) Give any LLM the whole documentation as Markdown. [shadcn registry, MCP and v0](https://docs.atomus.io/ai/registry/) Install Atomus components, screens and sections as source with shadcn add. [Agent compatibility (evals)](https://docs.atomus.io/ai/evals/) 50 prompts and six deterministic scorers, with and without the skill and MCP server. [Rules for agents](https://docs.atomus.io/ai/rules/) The core rules and the forbidden list every Atomus agent follows. --- # Claude Code, Cursor and Copilot > Set up coding agents to build UI with Atomus — the Atomus skill, the AGENTS.md template, docs for agents and the validator. URL: https://docs.atomus.io/ai/coding-agents/ Three pieces make a coding agent reliable with Atomus. The **skill** gives it the knowledge, the **AGENTS.md rules** make it follow it in your repo, and the **validator** catches what slips through. 1. **Install the Atomus skill** The skill follows the open [Agent Skills](https://agentskills.io/) format, so it works in Claude Code, Cursor, Copilot, Codex and other compatible agents. ```bash npx skills add StanVisionAgency/atomus ``` Or copy it by hand (Claude Code, all projects): ```bash git clone --depth 1 https://github.com/StanVisionAgency/atomus.git /tmp/atomus mkdir -p ~/.claude/skills && cp -R /tmp/atomus/skills/atomus ~/.claude/skills/ ``` For one project, copy it to `.claude/skills/atomus` in the repo and commit it, so the whole team gets it. 2. **Add the Atomus rules to your repo** Copy [`templates/consumer/AGENTS.atomus.md`](https://github.com/StanVisionAgency/atomus/blob/main/templates/consumer/AGENTS.atomus.md) into your product repo. Append it to `AGENTS.md`, or save it next to it and link it. Each tool reads it from a different place: Claude Code reads `CLAUDE.md`. Point it at the shared file: ```md See @AGENTS.md and @AGENTS.atomus.md. ``` Cursor reads `AGENTS.md` and `.cursor/rules/*.mdc`. To scope the rules to UI code, add a rule: ```md --- description: Atomus UI rules globs: src/**/*.{tsx,jsx,css} --- Follow AGENTS.atomus.md for every component, style and screen. ``` Copilot reads `.github/copilot-instructions.md` and `AGENTS.md`. Add a line to the instructions: ```md Build UI with Atomus and follow AGENTS.atomus.md. ``` 3. **Give the agent the docs** Point the agent at [`https://docs.atomus.io/llms.txt`](https://docs.atomus.io/ai/llms-txt/). Every page is also available as Markdown at `…/index.md`, for example `https://docs.atomus.io/components/button/index.md`. In Cursor, add `https://docs.atomus.io/llms-full.txt` under **Docs** and mention it with `@Docs`. 4. **Validate before you merge** The skill ships a dependency-free validator. Ask the agent to run it, or add it to CI: ```bash node ~/.claude/skills/atomus/scripts/validate.mjs src/ ``` ```text src/Settings.tsx:12:7 error invalid-value