basecoat-ui-mcp
Offline MCP server for Basecoat UI (shadcn/ui look, no React): templates, macro layouts, theming/dark-mode contract, static validation.
Install / Use
claude mcp add zygiu-zygis -- npx -y github:zygiu-zygis/basecoat-ui-mcpIf the server publishes to npm under a different name, use that package instead — check the repo README.
MCP Server
Model Context Protocol server
Quality Score
Category
LegalSupported Platforms
Our assessment of basecoat-ui-mcp
basecoat-ui-mcp scores 79/100 on our quality scale, 195th of 214 Legal skills we index.
Its MCP Server is 21 KB long, well organised into 25 sections with 14 code examples: a thorough specification that gives an agent plenty to work with.
It has 46 GitHub stars, so there is little community track record yet; judge it on its content.
Maintenance, license and trust
- The repository was last updated today, so basecoat-ui-mcp is actively maintained.
- It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
- Its trust signals score 97/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
basecoat-ui-mcp compared with similar skills
All 4 of these similar skills score higher than basecoat-ui-mcp; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| basecoat-ui-mcp (this skill)by zygiu-zygis | 79 | 46 | today | MCP Server |
| Agent-Reachby Panniantong | 100 | 91.8k | 20d ago | CLAUDE.md |
| headroomby headroomlabs-ai | 100 | 74.5k | today | CLAUDE.md |
| CowAgentby zhayujie | 100 | 47.2k | today | CLAUDE.md |
| Scraplingby D4Vinci | 100 | 85.9k | 1d ago | MCP Server |
Frequently asked questions
- How do I install basecoat-ui-mcp?
- Run
claude mcp add zygiu-zygis -- npx -y github:zygiu-zygis/basecoat-ui-mcp. The install tabs above show the steps for each supported agent. - Which AI agents does basecoat-ui-mcp work with?
- It is written for Claude Code, Claude Desktop, Cursor and Windsurf, as a MCP Server file. Other agents that read the same format can often use it too.
- Is basecoat-ui-mcp safe to use?
- It is MIT-licensed and scores 97/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is basecoat-ui-mcp still maintained?
- The repository was last updated today, so basecoat-ui-mcp is actively maintained.
Skill content
View source on GitHubThe server exposes a deterministic, offline registry for AI coding agents instead of requiring clients to scrape documentation. It runs over local stdio with zero runtime network requests, bounded < 2 KB responses, and static AST validation. The host application remains responsible for rendering, data access, authentication, sessions, credentials, OAuth, captcha, and other runtime behavior.
Install
Requires Node.js 22.14.0 or newer.
Quick Start by Client
Cursor (.cursor/mcp.json)
{
"mcpServers": {
"basecoat-ui": {
"command": "npx",
"args": ["-y", "@intellmedia/basecoat-ui-mcp", "--project-root", "${workspaceFolder}"]
}
}
}
Claude Desktop (claude_desktop_config.json)
{
"mcpServers": {
"basecoat-ui": {
"command": "npx",
"args": ["-y", "@intellmedia/basecoat-ui-mcp", "--project-root", "/absolute/path/to/your/application"]
}
}
}
Claude Code (CLI)
claude mcp add basecoat-ui -- npx -y @intellmedia/basecoat-ui-mcp --project-root $(pwd)
Generic npx (no install)
{
"mcpServers": {
"basecoat-ui": {
"command": "npx",
"args": ["-y", "@intellmedia/basecoat-ui-mcp", "--project-root", "/absolute/path/to/your/application"]
}
}
}
npm
npm install --global @intellmedia/basecoat-ui-mcp
Add the server to an MCP client. Use an absolute path for the host application so design context and persistent sessions belong to the intended project:
{
"mcpServers": {
"basecoat-ui": {
"command": "basecoat-ui-mcp",
"args": ["--project-root", "/absolute/path/to/your/application"]
}
}
}
The compiled entry can also be invoked directly:
{
"mcpServers": {
"basecoat-ui": {
"command": "node",
"args": [
"/absolute/path/to/node_modules/@intellmedia/basecoat-ui-mcp/dist/server/stdio.js",
"--project-root",
"/absolute/path/to/your/application"
]
}
}
}
Published packages include prebuilt dist/ and the immutable semantic snapshot
under src/semantics/. Server startup reads that snapshot and does not write
inside the installed package. The package contract test builds first, creates
and extracts an actual npm tarball, marks the extracted package tree read-only,
starts its dist/server/stdio.js under a network tripwire, completes MCP
initialize and tool-list requests, and closes the connection. A source checkout
does not include dist/.
From source
git clone https://github.com/zygiu-zygis/basecoat-ui-mcp.git
cd basecoat-ui-mcp
npm ci
npm run build
npm start -- --project-root /absolute/path/to/your/application
--project-root takes precedence over BASECOAT_PROJECT_ROOT, which takes precedence over the launch directory. The selected directory is the host project, not the MCP installation directory. The server exposes stdio only.
Project-root configuration
The configured root controls two things:
DESIGN.mdis read from that directory bybasecoat://project/context.- Persistent macro sessions are stored under
<project-root>/.basecoat/designer/. .basecoat/rhythm.jsonis an optional strict override. It must live under a real.basecoatdirectory inside the project root, be a regular non-symlink file no larger than 65,536 UTF-8 bytes, and resolve inside that root; missing or rejected files leave the packaged profile unchanged. Matching mapping IDs replace in place and keep packaged order.
Only that exact directory is used. The server does not walk parent directories. Keep .basecoat/ in the host project and back it up if design sessions are part of your workflow. The MCP server never writes host application source files.
Basecoat MCP tools
The server exposes three component tools, six macro tools, and two semantic tools. Every macro and semantic result is a bounded MCP packet of at most 1,999 UTF-8 bytes. Larger result sets use cursors; responses are not sliced mid-JSON.
Component tools
search_componentsreturns up to 8 compact{id, name, intent}summaries and never returns markup.intentandqueryare optional; an empty search returns an empty list.get_component_detailsreturns one Astro or HTML template with dependencies and composition guidance. Oversized entries fail closed.theme-toggleresolves totheme-switcher.validate_compositionstatically checks up to 256 KiB (262,144 UTF-8 bytes) of HTML or Astro source and returns at most 24 issues. Passcodeorhtml, and optionalsemanticProfileordensityProfile(comfortable|compact). It does not render, execute, resolve application modules, or certify accessibility.
Macro tools
The macro layer is this project's curated, project-specific composition system for shells, auth flows, data workspaces, slots, ports, rules, and recipes. It is compiled into a pinned registry snapshot. These layout contracts are not supplied automatically by shadcn or by MCP.
search_macro_blocks- find compatible blueprint blocks by query, role, family, or tag.get_macro_block- read a block section such asmanifest,structure,slots,ports,rules,dependencies, orprovenance.begin_design- create a persistent design session and pin its profile and registry revision. Pinned registries are hash-verified on load.get_design_context- read session lists or focused views such asoverview,graph,rules,focus, andnext.apply_design_patch- apply atomic graph changes withexpectedRevisionand an idempotentoperationId. A committed mutation always returns a success acknowledgement; oversized optional payload may be omitted withtruncated: true.validate_design- validate a draft or complete design graph and page its diagnostics.
Design sessions persist under <project-root>/.basecoat/designer/. Registry revisions are content-addressed; a session keeps using its pinned revision even after the package snapshot changes. Design snapshots use the form d:<designId>@<revision>.
Curated block blueprints:
- Shells & Navigation:
app-shell,sidebar-dashboard-shell,sidebar-inset-shell,sidebar-collapsible-icon,sidebar-mobile-flyout,page-header. - Authentication:
auth-sign-in,auth-sign-up,auth-split-screen. - Application & Workspace:
dashboard-workspace,dashboard-main,dashboard-activity,settings-workspace,data-table-detail-layout,detail-drawer-panel. - Forms & Data:
form-section,data-filters,data-records,data-pagination. - Marketing, Content & Utility:
pricing-tiers,newsletter-waitlist,empty-state,error-boundary,svg-area-chart,segmented-toggle.
Multi-block recipes:
workspace-dashboard- full analytics dashboard (KPI cards, activity chart, records table).workspace-settings- application shell with navigation, header, and settings workspace tabs.workspace-detail- data table workspace with slide-over detail inspection drawer.auth-flow- bidirectional authentication flow between sign-in and sign-up cards.auth-split-flow- split-screen auth frame pairing sign-in and sign-up with hero media.marketing-pricing- 3-tier pricing table paired with newsletter/waitlist banner.
Semantic tools
The semantic layer provides compiled rhythm profiles and finite state machine recipes:
get_rhythm_rules- access approved spacing, typography, surfaces, borders, and layout patterns by profile and family filters.get_fsm_recipe- read interaction states, events, transitions, guards, and actions fordialog,navigation(collapsible-navigation),auth-flow, andtabs.
Semantic tools use the same bounded pagination as macro tools. Each rhythm record keeps the semantic token ID separate from its approved Tailwind utility. Put the utility in class; keep the semantic ID in design metadata and agent reasoning. Project overrides from <project-root>/.basecoat/rhythm.json are reflected in an effective content ref and revision. FSM recipes are structural metadata, not runtime implementations.
End-to-end example
A dashboard page can follow this sequence:
get_design_context(view="sessions")
begin_design(designId="admin", profile="app-default", operationId="begin-admin")
get_rhythm_rules(profile="default", family="density")
search_macro_blocks(q="shell", limit=8)
apply_design_patch(
designId="admin",
expectedRevision=0,
operationId="add-dashboard",
operations=[{ op: "instantiate_recipe",
recipe: "workspace-dashboard",
pagePrefix: "admin" }]
)
get_design_context(view="overview", designId="admin")
get_macro_block(idOrRef="app-shell", section="structure", designId="admin")
get_fsm_recipe(recipe="dialog", section="transitions")
Implement the selected component leaves in the host project, then record completed regions with apply_design_patch(record_written). Finish with validate_design(mode="complete") and validate_composition on the generated Astro or HTML source. After a conflict or restart, reread the session and use the returned revision and cursors.
Copy templates/cursor/basecoat-designer.mdc into a host project's .cursor/rules for agent guidance. The MCP server does not write consumer application source files.
Auditing or restyling an existing UI
No design session is needed. Read basecoat://design/theming and basecoat://design/rhythm, call get_component_details for each control you change, then run validate_composition on the rendered layout source including <head> scripts and <style>. Theme rules:
theme-strategy-mixed-@media (prefers-color-scheme: dark)combined with the.darkclass strategy (dark tokens leak into light mode).theme-storage-key- a theme script persists under a key other than Basecoat'sthemeMode, creating a second resolver that drifts fromwindow.basecoat.theme.
Check light and dark screenshots with an OS dark preference and empty localStorage before deploying.
Source adaptation and offline import
The macro importer is intentionally local-first. It accepts a checked-out or otherwise
locally saved shadcn-style registry-item.json, a reviewed mapping, and a local
dependency lock:
npm run import:blocks -- \
--item=/path/to/registry-item.json \
--mapping=/path/to/mapping.json \
--lock=/path/to/dependency-lock.json \
--dry-run
This command does not fetch live shadcn registry items, install packages, execute upstream code, or make runtime network requests. Unknown, unlocked, or explicitly unsupported depe
Truncated for display — read the full file on GitHub.
Related Skills
Agent-Reach
91.8kGive your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
headroom
74.5kCompress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
CowAgent
47.2kOpen-source personal AI assistant & Agent Harness. Plans tasks, runs tools and skills, self-evolves with memory and knowledge. Multi-agent, multi-model, multi-channel. Lightweight, extensible, one-line install.
Scrapling
85.9k🕷️ An adaptive Web Scraping framework that handles everything from a single request to a full-scale crawl! Don't be shy, join here: https://discord.gg/EMgGbDceNQ and follow here for daily tips and tricks: https://x.com/Scrapling_dev
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
