Reference
Context Graph Glossary
The complete A–Z reference for context graph, decision trace, and AI agent infrastructure terminology. Every definition is written in the clear, concise format optimized for quick understanding and practical application.
Context Graph
A context graph is a living record of decision traces stitched across entities and time, so precedent becomes searchable. Unlike traditional databases that store what happened, context graphs capture why decisions were made—the exceptions, overrides, reasoning, and cross-system context that currently live in Slack threads, escalation calls, and people's heads. Context graphs are the data layer that enables AI agents to learn from institutional memory and make autonomous decisions with full traceability.
Learn more → What is a Context Graph?Decision Trace
A decision trace is a structured record that captures why a decision was made, not just what happened. It includes the specific context, reasoning, exceptions, approvals, and evidence that led to an outcome—forming the core data structure within a context graph. Decision traces enable AI agents to find relevant precedent when encountering similar situations, transforming isolated decisions into compounding organizational knowledge.
Learn more → Decision Traces ExplainedKnowledge Graph
A knowledge graph is a structured representation of entities and their static relationships, optimized for information retrieval. Knowledge graphs store facts about the world—such as “Company A is headquartered in New York” or “Product X belongs to Category Y”—but typically lack the temporal, causal, and decision-reasoning dimensions that context graphs provide. Knowledge graphs answer “What is X?” while context graphs answer “Why was X decided?”
Learn more → Context Graph vs Knowledge GraphContext Engineering
Context engineering is the discipline of designing and building dynamic systems that curate and maintain the optimal set of tokens presented to a large language model at inference time. Championed by Anthropic and practitioners across the industry, context engineering encompasses retrieval, tool calls, memory management, graph lookups, and compression strategies—controlling not just the instructions but the entire evidence environment the model reasons over. It has surpassed prompt engineering as the critical skill for building production AI agents.
Learn more → Context Engineering: The Critical DisciplinePrompt Engineering
Prompt engineering is the practice of crafting static instructions, examples, and formatting cues to elicit better responses from large language models. It focuses on the wording, structure, and arrangement of the prompt itself—such as few-shot examples, chain-of-thought instructions, or role-based framing. While prompt engineering remains a useful component within context engineering, it addresses only 5–10% of the tokens an agent sees; the other 90–95% comes from dynamically assembled context.
Learn more → Context Engineering vs Prompt EngineeringTribal Knowledge
Tribal knowledge is the unwritten institutional memory that exists only in the heads of experienced employees. It includes unwritten rules, exceptions, workarounds, relationship histories, and contextual understanding that people accumulate through years of experience but never formally document. Tribal knowledge is the single largest barrier to AI agent autonomy because agents cannot access what was never written down—context graphs solve this by capturing decision traces that encode tribal knowledge into a searchable structure.
Learn more → Tribal Knowledge and AIModel Context Protocol (MCP)
The Model Context Protocol (MCP) is an open standard originally developed by Anthropic that defines how AI models and agents consume contextual information from external sources. MCP provides a universal interface for connecting context producers (databases, knowledge graphs, APIs, enterprise systems) to context consumers (AI agents, coding assistants, chat interfaces), enabling a composable ecosystem where any context source can serve any agent. By 2026, MCP has become the de facto standard adopted by Glean, DataHub, Cursor, and dozens of other tools.
Learn more → Model Context Protocol (MCP) ExplainedGraphRAG
GraphRAG is a retrieval-augmented generation pattern that combines knowledge graph traversal with LLM generation. Instead of relying solely on vector similarity search to find relevant text chunks, GraphRAG uses structured graph queries to retrieve entity relationships, traverse multi-hop reasoning paths, and assemble context with full provenance. GraphRAG has been shown to dramatically reduce hallucination rates and enable complex reasoning that pure vector search cannot support.
Learn more → The Role of Graphs in Context EngineeringOntologyRAG
OntologyRAG is an ontology-driven approach to knowledge extraction and retrieval-augmented generation developed by TrustGraph. It uses formal ontologies as the organizing schema for what gets retrieved—rather than naively chunking documents and performing vector similarity search, OntologyRAG uses the ontology to understand what types of entities and relationships are relevant to a query, then traverses the graph to assemble precisely targeted context. This approach enables higher precision and more structured context assembly than standard RAG patterns.
Learn more → Ontologies and Context GraphsTemporal Context
Temporal context is the treatment of time as a first-class dimension in context graphs, enabling agents to reason not just about what is true now but about what was true at any point in the past. Temporal context captures when decisions were made, how knowledge evolved, what policies were in effect at specific moments, and how context shifted over time. It is essential for audit trails, compliance, and enabling AI agents to understand precedent within its proper historical setting.
Learn more → Temporal Context in Context GraphsBi-temporal Data Model
A bi-temporal data model is a data architecture that tracks two dimensions of time: event time (T)—when something was true in the real world—and ingestion time (T′)—when the system learned about it. This distinction is critical for enterprise AI because it enables agents to answer questions like “What did we believe to be true about this customer on January 15?” even if the data was later corrected. Zep's Graphiti temporal knowledge graph is the leading implementation of bi-temporal modeling for AI agent context.
Learn more → Temporal Context and Bi-temporal ModelsAgent Trace
An agent trace is an open standard for tracking AI code contributions, developed by Cognition AI (creators of Devin) and supported by Cursor, Cloudflare, and Vercel. Agent Trace captures the full context of how code is written, reviewed, deployed, and maintained by AI agents, creating a queryable graph of software development decisions. It represents a vertical-specific context graph standard that demonstrates how decision tracing applies to specific domains.
Learn more → Agent Trace: The Open StandardContext Engineering Patterns
Context engineering patterns are the four fundamental strategies for managing LLM context, identified by LangChain and widely adopted across the industry: (1) writing context—saving information outside the context window for later retrieval; (2) selectingcontext—pulling the right information in via RAG, graph queries, or tool calls; (3) compressing context—retaining only necessary tokens through summarization or distillation; and (4) isolating context—splitting context to help agents perform specialized sub-tasks without interference.
Learn more → Key Patterns of Context EngineeringInstitutional Intelligence
Institutional intelligence is the compounding organizational knowledge that emerges when context graphs capture every decision trace, enabling AI agents to learn from the collective experience of the entire organization. Unlike static knowledge bases that degrade over time, institutional intelligence grows stronger with every decision—each new trace adds precedent, refines understanding, and makes future AI-driven decisions more informed. It is the ultimate competitive moat for organizations deploying AI agents.
Learn more → Institutional Intelligence and AIAI Agent Memory
AI agent memory is the persistent context layer that enables AI agents to retain and recall information across interactions, sessions, and tasks. It goes beyond simple conversation history to include learned preferences, accumulated knowledge, relationship context, and decision precedent. Modern AI agent memory systems like Zep's Graphiti use temporal knowledge graphs to provide structured, time-aware memory that agents can query for relevant context at each step.
Learn more → AI Agent Memory SystemsSystems of Agents
Systems of agents is the emerging enterprise paradigm where autonomous AI agents replace traditional systems of record (CRMs, ERPs, data warehouses) as the primary interface for organizational operations. In this paradigm, the crown jewel is no longer the database—it is the context graph that enables agents to make decisions with institutional intelligence. Foundation Capital argues that the next trillion-dollar platforms will be systems of record for decisions, not data.
Learn more → Why Context Graphs Power Systems of AgentsContext Management
Context management is DataHub's term for an organization-wide capability to deliver the right data to AI context windows at the right time. It encompasses discovering what context exists across the organization, assessing its relevance to a given agent task, assembling it within token limits, and delivering it with appropriate provenance and governance metadata. Context management goes beyond traditional data management by focusing on the dynamic assembly of heterogeneous information for AI consumption.
Learn more → The Context Graph Ecosystem in 2026Guardian Agent
A guardian agent is Gartner's category for AI agents that monitor, govern, and align other AI agents to ensure they operate within organizational cultural and operational norms. Wayfound is the leading implementation of this concept, building a guardian agent that encodes company culture, risk tolerance, and communication styles into a context graph that serves as a contextual guardrail for other agents. Gartner predicts guardian agents could represent 15% of the AI agent market.
Learn more → Guardian Agents in the EcosystemContext Core
A context core is a versioned, portable context package developed by TrustGraph that bundles ontologies, extracted knowledge, and graph structures into a deployable unit. Context cores can be shared, versioned, and composed—enabling organizations to distribute curated context packages across teams, agents, and environments. They represent TrustGraph's approach to making context graph infrastructure modular and reusable, similar to how Docker containers made application deployment portable.
Learn more → Open-Source Context Graph ToolsEnterprise Graph
An enterprise graph is Glean's term for a combined knowledge graph and personal graph that maps both organizational information and the relationships between the people who create and use it. The enterprise graph fuses document knowledge (what information exists and how it connects) with social knowledge (who knows what, who works with whom, who authored what), creating a unified context layer that understands both the content and the human context surrounding it.
Learn more → Glean and the Enterprise GraphSearchable Precedent
Searchable precedent is decision history stored in a context graph that agents can query for guidance when encountering similar situations. When an AI agent faces an edge case—such as a pricing exception request or an unusual compliance scenario—it queries the context graph for similar past decisions: “How did we handle tier-1 customer escalations about pricing exceptions in Q4?” Searchable precedent transforms isolated decisions into compound intelligence that makes every future decision more informed.
Learn more → How Context Graphs Enable Precedent SearchEU AI Act
The EU AI Act is European regulation that requires organizations deploying high-risk AI systems to maintain auditable decision logs and demonstrate traceability of AI-driven decisions. It mandates that organizations can explain how and why their AI systems reached specific conclusions, making context graphs and decision traces not just a best practice but a regulatory requirement for companies operating in European markets. The EU AI Act is a key driver of enterprise investment in context graph infrastructure.
Learn more → AI Compliance and GovernanceContinue Learning
What is a Context Graph?
The definitive guide to understanding context graphs and how they power AI agents.
TECHNOLOGYContext Engineering
The critical discipline for building production AI agents in 2026.
INDUSTRY INSIGHTSThe Context Graph Ecosystem in 2026
Companies, tools, and the race to build AI's missing layer.
TOOLSOpen-Source Context Graph Tools
A guide to the open-source tools powering context graph infrastructure.