Enterprise Context Graph is the module where you model your business as a graph — the entities that matter (accounts, contracts, tickets, and so on), how they connect, and the real records that flow in from your connected systems. Agents, automations, and other platform features query this graph to reason about your business with context, instead of working off a single flat table.
Overview
A context graph is made of two things: node types, the entity types you care about (Account, Contract, Ticket, Employee, and so on), and edges, the relationships between them (an Account "has" a Contract, a Ticket "belongs to" an Account). Once you've shaped that schema, you connect structured and unstructured sources through a mapping wizard, and real records start flowing into your nodes and edges as those sources sync.
Every context graph is backed by a unified entity model behind the scenes, so its node types are really entities in that model — record quality, matching, and survivorship all run on that same layer. You never have to manage the model directly; it's simply what makes your graph's records reliable.
You can start a graph blank or from a curated template — a blueprint for a common domain like Customer Health, B2B Sales CRM, or Employee 360 — that arrives with its node types and edges already in place, ready to customize.
How the module is organized
Enterprise Context Graph opens on a list of every graph in your workspace. Opening a graph takes you into a workspace with five tabs, each covered in its own doc:
Graph list (Context Graphs Manager) — the table of graphs in your workspace, with status, version, and record counts; where you create, share, and delete graphs.
Graph (Context Graph Builder) — the schema canvas where you lay out node types and draw edges between them, and see how those edges become relationships between records.
Nodes (Context Graph Nodes) — where you define each node type's properties, primary key, colour, icon, and defalts, and see the deeper settings behind a node like survivorship and matching.
Sources (Context Graph Sources) — where you connect a data source, choose what to sync, optionally set up Discover & Enrich, and map source fields onto your nodes and edges.
Records (Context Graph Records) — where you browse, search, and inspect the actual records syncing into your graph, including drilling into how a record relates to others.
Each of these has its own page with full detail — this page is the starting point that ties them together.
How is this different from Ontology?
Ontology and Enterprise Context Graph both sit under AI Ontology & Data and rely on the same underlying entity model, but answer different questions. Ontology is where you define data models and let records flow through data quality, matching, and survivorship rules to produce a single trusted golden record per entity — your master data management layer. Enterprise Context Graph gives that same data shape as a graph you can query: node types and edges that agents, automations, and graph-query tools traverse to answer questions spanning multiple related entities, not just look up one record at a time.
In short: Ontology is about getting one clean record right. Enterprise Context Graph is about the connections between records, and about being the layer your AI agents and automations actually reason over.
Key concepts
| Concept | What it means |
| Graph | A named collection of node types, edges, and connected sources — the unit you create, publish, and share. |
| Node | An entity type in your graph, such as Account or Ticket, with its own properties and a primary key. |
| Edge | A relationship between two node types, such as an Account having a Contract. |
| Draft / Live status | Every graph starts as Draft. Publishing flips it to Live and stays Live — there's no unpublish. |
| Version | A change counter that increases every time the graph's schema changes; it isn't a count of how many times you've published. |
Notes
Publishing a source (which starts data syncing) and publishing a graph (which snapshots the schema) are two separate actions — doing one doesn't do the other.
Records are never snapshotted by a publish; they keep flowing into the graph continuously as sources sync, whether or not you've published.
There's currently no way to duplicate, archive, or pause a graph, and no picker to attach a graph directly to an AI agent — agents reach graph data through automations and tools that query it.
Whether Enterprise Context Graph appears in your sidebar depends on your workspace's navigation configuration; if you don't see it, ask your workspace admin.