Unsplash's own path costs you a developer account, an application and a key to rotate, for the sake of a hero image — and the shortcut everyone used is gone. Moving that registration server-side means an agent scaffolding a landing page can fill the images in the same pass as the markup. The batch manifest is the tool that makes that a single step rather than five.
A photo-fetching server that handles Unsplash's developer registration on its own hosted side, so the client needs no account. It exists because the zero-setup route everybody used, `source.unsplash.com/random`, was deprecated and now returns errors, leaving a lot of projects with broken image tags. The same engine is available as a CLI, an HTTP API and an embeddable widget.
- `search_images` — get image URLs for a query, with orientation, count and size constraints
- `save_image` — download one photo into a path in your project
- `search_and_save_image` — the two combined, for when you just want a file
- `batch_save_images` — fill an entire page from a JSON manifest of query, output path and orientation entries
- Orientation filtering across landscape, portrait, square or any, plus maximum width and height
- The same capability outside MCP: a CLI, a public HTTP endpoint with CORS enabled and no headers required, and a drop-in embed script
Nothing — no Unsplash app, no API key, no `.env` file. Install with `uv tool install tteg` or `pip install tteg` (0.4.2 in pyproject); that provides the `tteg-mcp` entry point, which your client runs over stdio. The README's client entry passes `tteg-mcp` to uvx as a package name, and PyPI has no package under that name. Result counts are capped between 1 and 10 per search, and the free tier allows 50 queries per day per IP.
One command — uv tool install tteg