AI and MCP
Textlab is written to be read by agents. Point your AI tool at it and it can find the right effect, read its props and install it.
Connect the MCP server
A read-only server over streamable HTTP. No key and no sign-in.
Claude Code
Terminal
claude mcp add --transport http textlab https://textlab.dev/api/mcpCursor, VS Code and others
mcp.json
{
"mcpServers": {
"textlab": { "url": "https://textlab.dev/api/mcp" }
}
}Clients that only speak stdio
Terminal
claude mcp add textlab -- node mcp/server.mjsThe stdio bridge in mcp/server.mjs forwards to the hosted server. Set TEXTLAB_URL to use a local copy.
Tools
search_components: find effects by intent, such as "a number that changes".list_components: browse everything, or one category.get_component: when to use it, props, variants, accessibility, motion notes, install command and the full source.get_install_command: one shadcn command for several components.get_tokens: the CSS tokens and motion tokens.
Add the agent skill
The Textlab skill teaches an agent when to reach for a component, how to install it and which rules to keep. Save it to .claude/skills/textlab/SKILL.md, or paste its rules into AGENTS.md.
Readable context
- llms.txt: every component with a one-line job and a link to its Markdown.
- llms-small.txt: one line per component, for small context windows.
- llms-full.txt: the whole library as one Markdown file.
- Any component as Markdown: add
.mdto its page URL, for example /components/scramble-text.md. - Registry index and catalog with metadata for every item.
Try a prompt
Prompt
Add a hero to the landing page with Textlab. Use the Textlab MCP server to find a title
reveal and an animated counter for the stats row. Install them with the shadcn CLI and
check the page with reduced motion turned on.