Skip to content
Image Asset Server

Image Asset Server quick start

Image Asset Server searches many free image sites at once and tells you, for every image, whether you can use it commercially. NASA, The Metropolitan Museum of Art, the Art Institute of Chicago, the Cleveland Museum of Art, the Library of Congress, Wikimedia Commons, Iconify and Openverse are searched directly; Unsplash, Pexels, Pixabay, Flickr, the Smithsonian, Europeana and the paid libraries (Shutterstock, Adobe Stock, Getty Images…) get a deep link with the same search. Results cover photos, public-domain art, maps and posters, illustrations, SVG vectors and icons, with pixel size, orientation and transparency where known, a thumbnail, a display-size preview and a licence verdict.

The public server at https://image.shep.bot is a free, non-commercial service and needs no account or API key.

Pick how you want to use it

You want to… Use Start here
Browse and download images yourself The web search /search
Pick every image a page, deck or post needs in one go The image set builder /build
Let Claude Code, Cursor, Copilot or another agent find images The MCP server at https://image.shep.bot/mcp Coding agents & MCP
Call it from a script, a static-site build or a backend The REST API REST API guide
Search and download from a terminal The CLI (npx image-asset-server) CLI
Explore every endpoint and try requests The interactive reference API reference

Connect your agent (30 seconds)

For Claude Code, run this once:

claude mcp add --transport http images https://image.shep.bot/mcp

Then ask for what you need:

Find a wide photo of a misty forest for our landing page hero that I can use
commercially, plus three SVG icons for upload, share and settings.
Tell me their licences and the credit lines I need.

Cursor, VS Code, Windsurf, Codex CLI, Gemini CLI and the Claude apps take the same URL. See Coding agents & MCP for each one.

Or call the API

# Landscape mountain photos, at least 1080 px on the short side, cleared for commercial use
curl "https://image.shep.bot/v1/search?q=mountain+lake&orientation=landscape&min_resolution=1080&commercial=true&limit=5"

# Details, every file size, the credit line and a display URL for one result
curl "https://image.shep.bot/v1/images/nasa%3APIA03606"

# Download a 1920 px copy (a redirect to the file on NASA's servers)
curl -L -o crab-nebula.jpg "https://image.shep.bot/v1/images/nasa%3APIA03606/download?width=1920"

What a search returns

Every search returns three things:

  • results: images merged from all sources and ranked by how well they match. Each has an id like met:436524, a category (nature.ocean, art.painting, archive.map, icons…), styles (photo, painting, vector, icon, transparent…), orientation, width, height, vector, transparent and year when known, a thumbnailUrl and a previewUrl that load in a browser straight from the source, a license with a verdict and an editorial flag, an attribution line when credit is required, and downloadable.
  • filters: what the server understood. Words like “4k”, “high resolution”, “vertical”, “png”, “svg” or “transparent” in the query become filters.
  • providers: one status line per source (ok, error, timeout, skipped or link). Sites that can’t be searched automatically come back as link with a searchUrl that runs the same search on their site.

A slow or broken source never fails the search.

No image passes through the server

Image Asset Server never proxies or re-hosts images. It searches the sources, reads their metadata, decides a licence verdict and writes credit files. Every pixel comes straight from the source’s servers:

  • Thumbnails and previews in the search grid load from the source (images-assets.nasa.gov, images.metmuseum.org, www.artic.edu/iiif, upload.wikimedia.org, api.iconify.design…).
  • The detail view shows the source’s file (the display URL from GET /v1/images/{id}).
  • Downloads (/v1/images/{id}/download) answer 302 with the file’s URL on the source.
  • Image sets (POST /v1/sets) return JSON, HTML and a download script whose file URLs point at the sources.

So the server stays small and cheap, the sources see real visitors, and nothing is redistributed.

Licences, in one line each

  • Commercial use OK (yes): use it, no credit needed.
  • Commercial use OK with credit (yes-with-credit): use it and credit the source with the attribution line.
  • Restricted, Needs a paid licence or releases, Unclear (restricted, clearance-required, check-terms): read the terms first.
  • Non-commercial only (no): study and personal projects, not ads or client work.
  • Editorial (license.editorial: true): people, insignia or brands without releases. Articles and education yes, advertising no.

commercial=true keeps only the first two verdicts. The full explanation is in Image licences explained.

For AI agents reading this

Machine-readable entry points:

Coding agents & MCP →

AI agent? This page is also available as markdown.