Memory

Memory is the layer that lets agents remember, and lets your platform know what an agent remembers and why. Every Rumbor Mission runs against a memory context that tracks provenance, trust, confidence, known time, and valid time for every fact it holds - not just a vector store of unstructured text.

Why it matters

Most agent stacks bolt memory on as an afterthought: a vector database, a summarization loop, and hope. That works until an agent needs to explain why it believes something, when it learned it, or whether a newer fact superseded an older one. Rumbor's memory layer is built to answer those questions from day one - every assertion carries the evidence trail behind it, queryable the same way you'd query any other governed system of record.

What it gives your agents

  • Provenance and trust - every stored fact tracks where it came from and how much to trust it, not just its content.
  • Bitemporal tracking - memory distinguishes when something became true (valid time) from when the system learned it (known time), so you can query "what did we believe on that date" as well as "what's true now."
  • Supersession, not silent overwrites - conflicting assertions don't quietly replace each other; memory tracks which fact superseded which, and why.
  • Queryable recall - agents (and your own tooling) retrieve memory through a real API, with results scoped to authorized context, not an opaque black box.

Using it from your own code

Rumbor Memory ships a public, provider-neutral Node.js/TypeScript client:

npm install @rumbor/memory-sdk
import { RumborMemory } from '@rumbor/memory-sdk';

const memory = new RumborMemory({
  baseUrl: 'https://your-rumbor-memory-host',
  apiKey: process.env.RUMBOR_API_KEY,
});

const result = await memory.recall({
  context: 'acme-prod',
  query: 'When does the team deploy?',
});

The SDK's full API reference - every method, type, and authentication detail - is published and kept automatically in sync with the SDK's source: rumbor-io.github.io/rumbor-memory-sdk.

Memory is one layer of the wider Rumbor platform - paired with Runtime, Gateway, Infrastructure, Identity and Policy, Observability, and Human Control, it's how Rumbor replaces the production-platform engineering every team eventually has to rebuild for their agentic products.