Glossary · Shared context graph

Shared context graph

A shared context graph is a body of machine-readable context that multiple AI agents contribute to and draw from, rather than each agent holding its own private memory. It lets context created by one agent be refined and reused by others, so knowledge accumulates across an organization instead of being rebuilt per assistant.

The problem it solves

Most agent memory is personal to one assistant and one session. That works for a chatbot and fails for an enterprise, where the same customer, contract, or asset is touched by many processes. Without shared context, the fiftieth agent starts roughly where the first one did.

How it differs from a database

A database stores records. A context graph stores relationships and the reasoning attached to them, including provenance for where each piece came from. That structure is what allows an agent to answer questions no single record contains.

Where it sits in the memory model

Shared context is the middle of three layers. Working memory is private to the agent that creates it, shared memory is where multiple agents coordinate, and verifiable memory is where context carries provenance others can independently check.

In practice

A procurement agent establishes that a supplier is dual-sourced through a plant in a sanctioned region. Under private memory that finding dies with the session. In a shared context graph it becomes context the compliance agent reads a week later, the logistics agent reads next quarter, and the audit agent cites the following year, each seeing where the finding came from and how confident it was.

FAQ

Common questions about shared context graph.

How is this different from RAG?
Retrieval augmented generation fetches documents to condition a single response. A shared context graph is persistent and multi-party: agents write to it as well as read from it, and what they write carries provenance that other agents can rely on.
Does sharing context mean exposing private data?
No. Working memory and private context stay in your infrastructure. What moves into shared memory is what you choose to share, with disclosure remaining under your control.
Why a graph rather than a document store?
Because enterprise questions are relational. Understanding whether a shipment can clear, or an asset is safe to run, depends on how entities connect, not on retrieving one document.
Related terms

Read next.

Start free

Power up your business with the OriginTrail DKG.

Create Knowledge Assets, build shared context graphs, enshrine decision traces. Open-source infrastructure. Community support. Zero cost.

Go Pro

The full Network Operating System.

Custom agent frameworks (Hermes, OpenClaw, LangChain, Claude). Custom data pipelines. Custom integrations with your enterprise systems. Dedicated infrastructure and support.