Skip to main content
Glama
noahchalifour

godly-mcp-server

README.md
# godly-mcp-server

An MCP server that lets AI agents search and view the UI/UX design inspiration curated on [godly.design](https://godly.design): landing pages, iOS apps, and website templates, with full-page and per-section screenshots.

It needs no API key. It reads the same public files the Godly site loads: `/search-data.json`, the Markdown versions of pages (`/websites.md`, `/website/<slug>.md`, ...), `/sitemap.xml`, and screenshots from `cdn.godly.design`.

## Tools

| Tool | Purpose |
| --- | --- |
| `search_inspiration` | Keyword search across websites, apps, and templates. Filter by `kind` and `category`, page with `limit`/`offset`. Leave out `query` to browse newest first. |
| `list_section_inspiration` | Browse curated `hero`, `cta`, `footer`, `og-images`, `logos`, or `trending` galleries. Each result includes a direct image URL for that section crop. |
| `list_categories` | Category names and counts for the `category` filter. |
| `get_inspiration` | Details for one item: description, live URL, metadata (industry, App Store rating, template price/author), and every screenshot URL. Set `include_preview: true` to also get the first screenshot inline. |
| `view_image` | Returns a `cdn.godly.design` screenshot as an image the model can see. Tall full-page captures are resized and split into vertical tiles; request more tiles with `page`. |

A typical agent flow: `search_inspiration` → `get_inspiration` → `view_image` (one tile per call).

## Setup

Requires Node.js 18.17 or later.

```bash
git clone https://github.com/noahchalifour/godly-mcp-server.git
cd godly-mcp-server
npm install      # also builds dist/
npm test         # end-to-end tests against live godly.design
```

### Claude Code

```bash
claude mcp add godly -- node /absolute/path/to/godly-mcp-server/dist/index.js
```

### Claude Desktop, Cursor, and other MCP clients

```json
{
  "mcpServers": {
    "godly": {
      "command": "node",
      "args": ["/absolute/path/to/godly-mcp-server/dist/index.js"]
    }
  }
}
```

## Notes and limits

- The site index is cached in memory for one hour.
- Godly publishes Markdown listings for about 40 recent websites and about 830 apps. Templates come from the sitemap, and only the newest ones carry titles and categories in the search index, so older templates match on their slug alone.
- Godly does not expose categories or styles for websites. Search websites by keyword (the description and industry are searched) or use the section galleries.
- Images are returned as JPEG at 1024px wide by default (configurable from 200 to 1280px), in tiles of up to 1600px tall, to keep each response a reasonable size for model context.
- This server depends on Godly's current page structure. If Godly changes it, the parsing in `src/godly.ts` may need updating.

## License

MIT

TDQS

A4/5.0

Scored across 5 tools

Disambiguation5/5

Each tool has a clearly distinct role: search_inspiration discovers slugs, get_inspiration retrieves full metadata, view_image renders screenshots, list_section_inspiration enumerates curated section galleries, and list_categories enumerates filter vocabulary. The descriptions explicitly chain the tools (search returns slugs -> pass to get_inspiration; section image -> pass to view_image), so there is no realistic misselection.

Naming Consistency4/5

All names follow a verb_noun snake_case pattern (get_, view_, search_, list_), which is predictable. The only blemish is list_section_inspiration, whose double-noun core is slightly awkward compared to sibling names like list_categories.

Tool Count5/5

Five tools is well within the ideal range and every tool maps to a distinct step in the browse-inspect-view workflow. Nothing feels padded or redundant for a read-only inspiration gallery.

Completeness4/5

The read-only lifecycle is covered end to end: discover (search/browse/sections/categories), inspect (get_inspiration), and view (view_image with tiling). Minor gaps exist — no explicit platform or industry filter exposed as a tool, and no related/similar-item lookup — but agents can work around these via search_inspiration.

Maintenance

ActivityMaintained
ResponsivenessNo issues