HasData Google Images

Google Images MCP Server

A hosted Google Images MCP server that gives Claude, Cursor, Windsurf and any other MCP client one read-only Google Images tool. Run an image search with size, colour and type filters and get 100 results a call, each with the page it came from and the full-resolution file URL, as structured JSON, with nothing to host.

It reads the Google Images results page a signed-out visitor sees.

1,000 free credits every month, no card required, which is 100 Google Images calls at the 10-credit rate.

https://mcp.hasdata.com/mcp?apis=google_images

• Prompts and resources

Contents

• What you need
• Quick start
• Example prompts
• Tools
• Errors and failure paths
• Pricing, free tier and limits
• How it compares
• FAQ
• HasData links
• Development
• Contributing
• License

What you need

An MCP client and a HasData API key from the dashboard, free to create with no card, and the free tier covers about 100 calls a month at the 10-credit rate. This is a remote server, so the simplest path is a URL and an x-api-key header, with no container to run. A client that only speaks stdio reaches it through a thin launcher, published as @hasdata/google-images-mcp on npm and hasdata-google-images-mcp on PyPI, shown below.

Quick start

The server URL is the same for every client. We run it hands-on in Claude Code and Claude Desktop. The other blocks follow each client's own documented format for a remote server.

Clients with OAuth support can add the same URL as a connector and sign in without putting a key in a config file.

Claude Code

claude mcp add --transport http google-images "https://mcp.hasdata.com/mcp?apis=google_images" \

--header "x-api-key: HASDATA_API_KEY"

Claude Desktop

Settings, then Connectors, then Add custom connector, then paste https://mcp.hasdata.com/mcp?apis=google_images and sign in.

For the config-file route, Claude Desktop loads only local (stdio) servers, so it reaches a remote server through a stdio launcher. The @hasdata/google-images-mcp package is that launcher, and it reads the key from the environment. Add this to claude_desktop_config.json:

{

"mcpServers": { "google-images": { "command": "npx", "args": ["-y", "@hasdata/google-images-mcp"], "env": { "HASDATA_API_KEY": "YOUR_KEY" } } } }

For Python instead of Node, swap the launcher for the PyPI package, which uvx runs without a manual install:

{

"mcpServers": { "google-images": { "command": "uvx", "args": ["hasdata-google-images-mcp"], "env": { "HASDATA_API_KEY": "YOUR_KEY" } } } }

Cursor

~/.cursor/mcp.json for every project, or .cursor/mcp.json for one:

{

"mcpServers": { "google-images": { "url": "https://mcp.hasdata.com/mcp?apis=google_images", "headers": { "x-api-key": "HASDATA_API_KEY" } } } }

Windsurf

~/.codeium/windsurf/mcp_config.json. Windsurf calls the field serverUrl, not url:

{

"mcpServers": { "google-images": { "serverUrl": "https://mcp.hasdata.com/mcp?apis=google_images", "headers": { "x-api-key": "HASDATA_API_KEY" } } } }

VS Code

.vscode/mcp.json in the workspace:

{

"servers": { "google-images": { "type": "http", "url": "https://mcp.hasdata.com/mcp?apis=google_images", "headers": { "x-api-key": "HASDATA_API_KEY" } } } }

Example prompts

• Find images of the Austin skyline and give me the ones over 1500 pixels wide.
• Which sites are ranking images for this product name?
• Get black and white photographs for this query, skipping clipart and line art.
• Search this query from the German Google and compare the sources with the US ones.
• Find images of this landmark and tell me which are stock photos.
• Pull the next page of results for this query.

One call answers each of these, because a page carries 100 results. Paging is only needed past that.

Tools

One tool, 10 credits per successful call.

Get image search results

hasdata_google_images_images_getImageSearchResults

One page of Google Images results.

Returns an imagesResults array. Each entry carries position, title, link to the page hosting the image, source as the site name, thumbnail on Google's CDN, original as the full-resolution file, and originalWidth and originalHeight.

{

"position": 1, "title": "The city skyline of Austin, Texas - Backyard Image", "link": "https://www.backyardimage.com/the-city-skyline-of-austin-texas/", "source": "Backyard Image", "thumbnail": "https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcSOGtSVdgWtQtW1_x8-gi7iQckNHXBQ5FdKqEpCKXsbWg&s=10", "original": "https://www.backyardimage.com/wp-content/uploads/2025/01/2025-01-12-drone-Heap-0015-Pano-Edit.jpg", "originalWidth": 1200, "originalHeight": 615 }

The tbs parameter is Google's own filter string, and several filters combine with commas.

Prompts and resources

The server exposes 5 resources, one per parameter whose accepted values are a fixed list. Reading one is cheaper than learning the vocabulary from a rejected call, and it costs no credits. Each URI is hasdata://google_images/ .

The list is served without an API key, so a client can read it before a user has signed up.

Errors and failure paths

Plan for these rather than assuming a happy path.

`original` is a third-party URL and it can be dead, hotlink-blocked or slow. It points at the file on whichever site hosts it, so it carries none of Google's availability. Fetch it defensively, and fall back to thumbnail, which is served from Google's CDN.

`thumbnail` is a Google CDN link, not an embedded image. It is a real URL rather than a data URI, so displaying one still costs a request to gstatic.com.

A page carries 100 results, not 10. Cost is per call rather than per result, so one call covering 100 images is the cheapest shape here. Reach for ijn only past the first hundred.

`ijn` starts at zero. Passing 1 gets the second page, not the first.

Dimensions describe the original, not the thumbnail. originalWidth and originalHeight are the file behind original, so a size filter applied locally has to read those rather than measure what you displayed.

`tbs` values are Google's, and an unrecognised one is ignored rather than rejected. A filter that silently does nothing looks the same as a filter that found everything, so check that the results actually narrowed before trusting a tbs string you assembled by hand.

`domain`, `gl` and `hl` are three separate axes. The domain is which Google answers, the country code is where the search comes from, and the language is the interface. Changing one moves the results less than you expect.

Results that carry data also carry a requestMetadata.id worth quoting in support.

Pricing, free tier and limits

The Google Images tool costs 10 credits per successful call. Response size does not change the price, and a call returns 100 results, which makes this one of the cheaper tools in the catalogue per row returned.

The free tier is 1,000 credits every month with no card, which is 100 Google Images calls at the base rate. It renews with the billing cycle, so a low-volume agent runs on the free tier indefinitely.

Paid plans start at $59 a month for 200,000 credits, which is 20,000 calls. The unit price falls with volume, from $1.48 per 1,000 calls on the entry plan to $0.60 on Basic and $0.41 across the Growth tiers. Current figures live on the pricing page.

Your plan also sets concurrency. The free tier allows 5 requests at a time, Startup 5, Basic 15, and the Growth tiers run from 50 to 500. Retry on the 429 with a backoff in anything unattended, because an agent that sweeps a keyword list will reach the ceiling before you do.

A re