Webfetch

What ashlrai/webfetch MCP server does

The ashlrai/webfetch MCP server gives MCP-compatible agents a single interface for searching image sources that expose structured metadata. It federates requests across 25 providers, including Wikimedia, Openverse, Unsplash, Pexels, Pixabay, iTunes, MusicBrainz, Flickr, Internet Archive, NASA, Smithsonian, and other cultural or image-search services listed by the project.

Results include a provider, image dimensions, a license classification, attribution text, and a URL suitable for downloading. The ranking policy favors known license tags and stronger metadata before resolution and provider priority. Results marked UNKNOWN are excluded by default, which suits workflows that need a defensible licensing trail rather than maximum search coverage.

How it works

The MCP surface accepts structured calls from an MCP client. The documented examples include search_images for a general query, search_artist_images for artist-related images, and download_image for retrieving an image from a result URL. Artist searches can include a type such as portrait.

Local mode calls the webfetch core library in the same environment as the server and uses provider credentials from environment variables when a selected provider needs them. Providers without credentials can still be used when their APIs are available without authentication. The project also describes a cloud mode for hosted API access, pooled keys, managed browser fallback, usage accounting, and team controls; that mode uses the webfetch cloud API and requires its own API key.

Setup and configuration

Run the MCP server with:

npx -y getwebfetch-mcp

The repository lists Cursor and Cline among supported MCP clients, along with Claude Code, Continue, Roo Code, and Codex. Client configuration details are not included in the provided material beyond the MCP entry point.

Provider credentials are optional at the overall server level but may be needed for specific sources. Examples include UNSPLASH_ACCESS_KEY, PEXELS_API_KEY, PIXABAY_API_KEY, BRAVE_API_KEY, BING_API_KEY, SERPAPI_KEY, FLICKR_API_KEY, and Spotify client credentials. The cloud path uses WEBFETCH_API_KEY; local or cloud selection can also be controlled with WEBFETCH_MODE.

Tools and capabilities

The ashlrai/webfetch MCP server documents these MCP operations:

• search_images({ query, limit }) searches across the configured provider set.
• search_artist_images({ artist, kind }) targets artist imagery such as portraits.
• download_image({ url }) downloads a selected image.

The broader project also provides CLI operations for image search, artist search, album search, downloads, probing, license inspection, and batch processing, but those are separate CLI surfaces rather than additional MCP tools confirmed in the excerpt. The repository describes a local cache for repeated downloads and exposes attribution information for selected results.

Limitations and notes

License labels describe the source metadata and platform terms; they do not replace checking the applicable provider terms for a particular use. Unsplash, Pexels, and Pixabay are represented with platform-specific tags rather than being treated as Creative Commons or CC0. General web-search providers may return UNKNOWN classifications and can be opt-in or require API credentials.

The browser and managed-browser providers are fallback options and use UNKNOWN licensing by default. Their availability can depend on configuration, provider limits, and, for managed browsing, a Bright Data token. Cloud mode is separate from the local default and sends requests to the hosted webfetch API.