Neo4j Releases Relate: An Open-Source Embedded Database for AI Agents
Graph database leader Neo4j has introduced Relate, an AI-native, open-source in-memory embedded database built in Rust for AI agents and agentic systems, offeri

On November 11, 2026, graph database provider Neo4j officially announced the release of Relate, an AI-native, open-source embedded database built specifically for AI agents and autonomous agent systems. Operating in-memory and in-process, Relate is engineered to meet the high-frequency read and write demands of modern agentic workflows.
As autonomous systems increasingly surpass human conversational assistants in enterprise workloads—projected by Gartner to reach 75% of database management system workloads by 2029—infrastructure requirements have shifted toward lightning-fast, resource-efficient data stores. Relate addresses this shift by allowing developers and AI coding agents to provision a database and begin building applications within milliseconds.
Rust-Based Architecture and Sub-Millisecond Execution
Relate is built from the ground up in Rust, a systems programming language known for memory safety and high execution speed. This foundation enables the database to deliver sub-millisecond read performance in memory.
- In-Process and In-Memory Operation: Relate runs directly inside the agent's process space without requiring a separate database server, minimizing network latency and overhead.
- Durable Storage Support: While operating as a lightweight in-memory store, it includes optional fsync'd write-ahead logging (WAL) and disk-flush capabilities to ensure long-term data persistence when needed.
- Standard Query Compliance: The database provides 100% openCypher and over 97% GQL (Graph Query Language) compatibility out of the box.
Native Integration Built for Autonomous Agents
Enterprise budgets are rapidly pivoting from simple conversational retrieval tools toward advanced autonomous systems capable of planning, executing, and coordinating multi-step operations. Relate is engineered specifically for these execution loops.
- Multi-Hop Reasoning and GraphRAG: It brings multi-hop reasoning capabilities and GraphRAG (Graph-based Retrieval-Augmented Generation) directly to AI agents, reducing hallucinations and improving factual accuracy.
- Automatic Harness Detection: Relate natively detects agent harnesses and returns structured errors equipped with explicit retryable flags, enabling agents to handle exceptions and self-correct without custom parsing logic.
- Context Window Protection: To prevent context window saturation during large retrievals, Relate enforces a row cap combined with an in-band truncation marker, ensuring that agents receive bounded and predictable payloads.
Seamless Integration Across Modern AI Stacks
Relate deploys seamlessly wherever AI agents operate, whether inside Jupyter notebooks, web browsers, software containers, or edge devices.
- Out-of-the-Box Connectivity: The database ships with built-in agent connectivity via a Model Context Protocol (MCP) Server and a dedicated command-line interface (CLI) for direct execution.
- Standard Endpoints: It exposes Bolt and HTTP Query API v2 endpoints to provide drop-in support for existing graph drivers, orchestration frameworks, and developer tooling.
Released under the open-source Apache 2.0 license, Relate is available for installation via its official portal, offering developers a purpose-built data layer for agent memory and sandboxed execution environments.
Sources
- HPCwire: Neo4j Releases Relate Open Source Database for AI Agents
- Neo4j Official Website: Relate by Neo4j