Working with Context

Ingest data, shape the Context Graph, and retrieve context for your agent

Zep turns the data you send into agent memory: a temporal Context Graph of entities, relationships, and facts, then a token-efficient context string for your agent’s prompt.

Work through the stages in order, or jump to the one you need.

StageWhat you doStart here
IngestAdd chat, business data, documents, or JSON so Zep can build the user’s Context GraphIngest
Shape the GraphCustomize extraction — ontology, domain terminology, summaries, and observationsShape the Graph
RetrieveAssemble facts, entities, summaries, and other context types into a prompt-ready stringRetrieve

Only Ingest and Retrieve are required. Zep extracts entities, relationships, and summaries automatically, so shaping the graph is optional — reach for it when your domain needs a specific ontology or terminology.

Graph search, reads, and CRUD sit outside this lifecycle under Working with Graphs, starting with searching the graph.