Skip to main content
Use the Cache plugin to cache tool results and control TTL/refresh behavior. Overview
  • Opt in per tool via metadata (e.g., cache: true or { ttl, slideWindow }).
  • Configure the store at app level: in‑memory or Redis (client or config).
  • Cache keys include tool id and validated input; you can extend via hooks.
Set up (pseudocode) 
import CachePlugin from '@frontmcp/plugins/cache'

@App({
    plugins: [CachePlugin], // or CachePlugin.init({ defaultTTL: 300 })
})
class MyApp {}
Opt in per tool (pseudocode) 
export const GetReport = tool({
   id: 'get-report',
   inputSchema: { id: z.string() },
   cache: { ttl: 600, slideWindow: true },
})(async (input, session) => {
   // expensive computation or remote call
   return await fetchReport(input.id);
});
Notes
  • Memory store is per‑process; prefer Redis in production.
  • slideWindow updates the TTL on access; without it, TTL is absolute.
  • Hooks can decorate the key or short‑circuit on policy.