YouTube MCP for Claude & ChatGPT
TubeLab is the YouTube MCP Server for Claude and ChatGPT. Real channels, viral outliers, video transcripts and much more - all in your AI chat.
翻译准备中
尚未检测到更新。GitHub 发布版本后会显示在这里。
Design review skill for UI code: accessibility (WCAG 2.1) + visual design checks with concrete fixes. From rams.ai
翻译准备中
正在加载完整 README。
查看完整 README ↗A design review for UI code the way a senior designer would give one: craft and visual quality first, accessibility built in (WCAG 2.2), with concrete fixes. Made by rams.ai.
As a Claude Code plugin (this repo is its own marketplace):
claude plugin marketplace add rams-design/rams-plugin
claude plugin install rams@rams
Once Anthropic's directory lists Rams, you can also add it from claude.ai/directory, and it follows you into Claude Code.
Via the open skills ecosystem (skills.sh, installs into 20+ agents):
npx skills add rams-design/rams-plugin
As a Codex plugin (this repo is its own marketplace):
codex plugin marketplace add rams-design/rams-plugin
codex plugin install rams@rams
The ChatGPT and Codex plugin directory takes a ZIP: scripts/build-openai-zip.sh
builds it from .codex-plugin/plugin.json.
Or the one-line installer (Claude Code, Cursor, Codex, Windsurf, Amp, OpenCode, Gemini CLI):
curl -fsSL https://rams.ai/install | bash
Ask for a review naturally ("review this component for design issues") or run
/rams path/to/Component.tsx. The review covers accessibility (alt text,
labels, keyboard access, focus, contrast, touch targets) and visual design
(spacing, typography, color, component states), each finding with a fix.
Without the hosted engine, this plugin runs locally as a heuristic review. The hosted engine at rams.ai reviews with the full scored rule set (live list at rams.ai/rules), verified re-reviews, and one-click fix suggestions.
The plugin also bundles the Rams MCP, the hosted engine as tools your agent calls:
quick_review: the check to run while editing. Top issues with severity,
category and file:line in about ten seconds.review_files: before committing. A 0-100 score and patches your agent can
apply, the same judgment a pull request gets.verify_fixes: after applying fixes. Confirms they landed, and is free.review_url: reviews a live public page by its address, at desktop and
phone width. Top issues and a score in up to a minute, one review credit, no
patches.usage: reviews used and left on your plan. Free.To connect it in Claude Code, run /mcp, pick rams, and approve access in
the browser. You sign in or create a free account as part of that step, and
there is no key to paste. The free plan includes 15 reviews to try it. A quick
check costs a fifth of a review on paid plans and half on the free plan.
Without it connected, the local skill still works as a checklist review.
/rams command are plain Markdown instructions. They run
nothing on their own and send nothing anywhere.https://worker.rams.ai/mcp. Nothing is
installed or run on your machine for it.quick_review, review_files or verify_fixes, it
sends the contents of the files it chose to review to that server, signed in
with your Rams account. review_url sends only the address of the public
page to review. Rams reviews them and sends back findings, a score
and patches. Nothing is sent until you connect the server and a tool is
called.MIT licensed.