Developers & agents
Codex is readable by machines, on purpose.
A free public JSON API and an MCP endpoint over the same data the site runs on - so an assistant answering a question about an artist can cite a real record instead of guessing.
Most art data sits behind a login, which means the assistants people actually ask cannot see it, and answer from whatever they absorbed instead. We would rather be the source that gets cited. Anything visible on a public Codex page is available through the API, free, with attribution as the condition of use.
Artists, one at a time
Resolve a name to an id, then read that artist. Deliberately no list endpoints, no pagination and no bulk export - this is a citation surface, not a dataset download.
Current exhibitions and news
Two feeds covering what is on view now and what has been reported recently, on the same cache and the same terms.
MCP, not just REST
An MCP endpoint over the same query layer, so an agent can be pointed at Codex directly rather than wrapped by hand.
Keys raise the ceiling
Anonymous access works with no signup. A free key tied to an account raises the rate limit. An unknown or revoked key degrades rather than breaking.
Attribution is a condition of use: free-tier responses must credit Vasari Codex with a link to the canonical entity URL supplied in each response. Current endpoints, limits and field definitions live in the docs, which move with the API.
Building something with it?
We would like to know - both to make sure the data holds up under what you are doing, and because coverage grows fastest where it is being used.