REST Cache

Appearance

Introduction ​

REST Cache stores the responses your Strapi API sends, and serves them again without touching the database. When content changes, it clears what that change affected.

The second half is the difficult half, and it is what this plugin is really for. Caching is easy; knowing when to stop is not.

How it works ​

A GET to a cached route is answered from the store when an entry exists for its cache key, and otherwise passed through, with the response stored on the way back.

Invalidation hooks the document service, which every write in Strapi passes through — REST, GraphQL, the content manager, the deprecated entity service, scheduled Content Releases, review workflows, and any strapi.documents() call of your own. A change made through any of them clears the entries it affects, including those of related content types.

That matters because the obvious alternative — watching HTTP routes for writes — cannot see a GraphQL mutation or a scheduled release, and drifts out of step with Strapi's route list every time that list changes.

What you get ​

  • Pluggable storage. In-memory for a single instance, Redis (or KeyDB, or Valkey) when several instances need to share one cache, or your own.
  • Per-content-type and per-route control over lifetime, key composition and when to bypass the cache entirely.
  • Cache keys you decide. Query parameters, request headers, and per-caller keys for authenticated traffic.
  • Request coalescing. Concurrent requests for the same missing key make one call to the origin, not one each — which is exactly what you want on a cold start or straight after a purge.
  • ETag and 304 support, and X-Cache headers for seeing what happened.
  • An admin panel showing what is cached right now, with purge controls.

When not to use it ​

This plugin caches REST responses. It does not cache GraphQL responses — though a GraphQL mutation still invalidates what it changes, so the two coexist safely.

It also will not make a slow query fast for the person who triggers it. A cache miss costs the original query plus the write to the store. What it changes is what the next reader pays.

Start here ​