Neo4j on November 11 released Relate, an open-source, in-memory, in-process embedded database built in Rust and engineered for AI agents and agentic systems1,2. Licensed under Apache 2.0, Relate is designed to let AI coding agents create a database and begin building applications in milliseconds.
Architecture and performance
Relate delivers sub-millisecond reads in memory to handle the high-frequency read/write cycles characteristic of agentic workloads. It maintains transactional integrity through an in-memory engine with an optional fsync'd write-ahead log for durable stores, and provides native options to flush state to disk for long-term persistence. A single Cypher statement can search by vector similarity and execute graph traversals simultaneously, bringing multihop reasoning and GraphRAG capabilities to agents for what Neo4j describes as more accurate results and fewer hallucinations.
Relate is more than 97% GQL compliant and 100% openCypher compatible out of the box. It exposes Bolt and HTTP Query API v2 endpoints, providing drop-in support for existing graph drivers, orchestration frameworks, and standard developer tooling.
Agent-native design choices
The database ships with built-in agent connectivity via a Model Context Protocol (MCP) Server and a dedicated CLI for direct execution. It natively auto-detects agent harnesses and returns structured errors with explicit retryable flags. To prevent context-window saturation during large retrievals, Relate enforces a row cap combined with an in-band truncation marker.
Relate deploys across Jupyter notebooks, web browsers, software containers, and edge devices, and can run in-process for agent sandboxes and agent memory as well as on mobile or edge hardware.
Market framing
Neo4j cited a June 2026 Gartner report, "Forecast Analysis: Agentic AI in DBMS, Worldwide, 2026," which projects that machine users will surpass human and AI assistant usage by 2029, reaching 75% of DBMS workloads, up from 10% in 2025. The same Gartner analysis noted that AI agents are generating 10 to 100 times more database interactions than humans and driving concurrent user loads upward.
ANALYSIS By packaging graph traversal, vector search, and an MCP server into a single embeddable binary under a permissive license, Neo4j is positioning Relate as infrastructure that agent frameworks can adopt without external database dependencies or licensing friction.