Neo4j releases Context Engine, a knowledge layer for AI agents
An analyst says that Context Engine is read only, but Neo4j says it gives agents a map of company data.
Claimed, not confirmed
On Tuesday, Neo4j, a graph database company, released Context Engine. It is a set of tools that helps customers make a knowledge layer. This is a map of business data that agents use to answer queries. Neo4j says that many customers could not make agents read account and supplier the same across systems. Neo4j gave no test results.
What is in it
- Business Ontology: a model that maps data in Neo4j.
- Context Build and Runtime: make ontologies from sources such as Databricks and Snowflake.
- Context Studio: control the ontologies.
- Context Memory: memory that stays for agents.
The proof
Neo4j gave no test results or benchmark numbers. William McKnight of McKnight Consulting says that it is different from rivals because it runs queries in its database. McKnight says that Neo4j must publish benchmarks from an external group.
What is missing
McKnight says that Context Engine is read only. It has no connectors that copy changes from other systems. Agents cannot write changes back to the systems of record of a company.
Insight: Vendors compete to own the layer under agents
Kong says its Code Mode lets an agent find operations at runtime. AWS released a sandbox with a policy language to set the tasks of agents. Neo4j offers the data map. This reads as a move to own the layer between agents and data. The first layer a company adopts can set the standard.
Sources
Posted
