Memory your agent is actually allowed to use.
The Agent Memory SDK governs what an AI agent may recall and act on. Every memory read is authority-checked at the moment of recall; every recalled fact carries a content hash and a signed provenance record — so a decision built on memory is replay-provable, not hand-wavy. TypeScript · Python.
Recall is an authority act
A clean prompt can still spend authority on a fact it was never allowed to remember. This SDK treats every recall as a governed action, not a free vector lookup.
Authority-checked recall
Each memory read resolves against purpose + scope before the fact reaches the model. Out-of-scope memory is withheld, fail-closed.
Content-hashed provenance
Every stored and recalled item carries a content hash + signed provenance, so an auditor can prove exactly which memory drove a decision.
Quarantine & revocation
Memory capabilities can be quarantined or revoked; the engine emits the governance event family on every state change.
Install
Backed by the KYE Memory Engine™. The wire contract + verification surface are open.
# TypeScript / JavaScript npm install @kye/agent-memory # Python pip install kye-agent-memory
import { MemoryClient } from "@kye/agent-memory"; const mem = new MemoryClient({ trustDomain: "acme" }); // Recall is authority-checked at the moment of read const recall = await mem.recall({ agent: "support-bot", purpose: "resolve-ticket", query: "customer payment history", }); // recall.items each carry contentHash + signed provenance // out-of-scope memory is withheld, not returned
Where the SDK sits in the lifecycle
The Agent Memory SDK governs the recall surface that feeds Authority Resolution in the KYE Agentic Governance Lifecycle™.