What is a context graph? And where your customers fit in it.
A context graph is a connected model of a company that AI agents can query: the people, systems, documents and decisions, how they relate, and how they change over time. It gives an agent the “how” and “why” behind the work, not just the files.
The short definition
A knowledge graph stores things and how they connect: this customer, that ticket, this owner. A context graph adds what an agent needs to act: time (what was true when), history (who decided what, and why), and rules (who is allowed to see or change it). Glean describes it as modeling how work actually happens, built from traces of actions across your apps (Glean). Atlan describes it as data assets, their relationships and the history of decisions made about them (Atlan).
What goes into a context graph
🧩 Entities
🔗 Relationships
⏳ Time
⚖️ Decisions and rules
Context graph vs knowledge graph vs context layer
| 🕸️ Knowledge graph | 🧠 Context graph | 🗂️ Context layer | |
|---|---|---|---|
| 📦 What it holds | Entities and links | Entities, links, time, decisions, rules | Whatever context the company serves to people and agents |
| 🛠️ Built for | Search and lookup | Agents that act over long tasks | Every AI tool in the company, usually over MCP |
| ❓ Typical question | "Who owns this account?" | "How do escalations usually get resolved here, and why?" | "What should this agent know before it starts?" |
A context graph is one way to build a context layer. Our guide to the product context layer covers the product team's slice.
The part most context graphs miss: the customer
Look at what goes in: documents, tickets, data assets, process traces. These are things your company writes down or does. Customer needs are different. They arrive every day in sales calls, support tickets, interviews and surveys, in the customer's own words, and nobody writes them down as a tidy record.
So when an agent asks a context graph “How many customers asked for this feature last quarter, and what are they worth?”, it can find tickets that mention it. It can't count distinct customers, because nobody grouped the needs, removed duplicates or tied them to accounts.
What a customer layer adds to the graph
🎯 Needs as nodes
💰 Edges to accounts and revenue
💬 A quote on every edge
⏳ Time
Retrieval finds documents. Counting finds the roadmap. That's what Evermuse adds to your context graph or company brain: counted, cited customer needs, served to any agent over MCP.
FAQ
What is a context graph?
A connected model of a company's entities, relationships, history and rules that AI agents query at runtime, so they understand how work happens and why, not just what files exist.
How is a context graph different from a knowledge graph?
A knowledge graph stores entities and links. A context graph adds time, decision history and policies, which agents need to act safely.
Does a context graph include customer feedback?
Usually only as documents or tickets. Counting how many customers raised a need, and tying it to accounts and revenue, takes a customer layer that processes every conversation ahead of time.
Add the customer to your context graph.
Connect your calls, tickets and chats. Evermuse counts every request, ties it to the account, and links every answer to its source.