AboutAI ToolsCursorClaude CodeLovableBolt.newAI tool reviewsTutorialsVibe codingAI agentsMobile apps with AIAll tutorialsBlogSponsorWork With MeJoin 285K+ Subscribers

12 November 2026 · 3 min read

Neo4j Launches Relate: An Open-Source Embedded Graph Database for AI Agents

Neo4j has released Relate, an open-source embedded database built in Rust for AI agents. It runs in-process with optional disk persistence and includes graph-native features for agent memory, sandboxes, and edge deployments.

What was announced

On 11 November 2026, at its NODES 2026 conference, Neo4j released Relate, an open-source embedded database purpose-built for AI agents and agentic systems [2]. The database is licensed under Apache 2.0 and written in Rust. It runs in memory inside an application’s own process, with optional disk persistence when durable storage is needed [1].

Relate is designed to let AI coding agents create a database and start building applications in milliseconds. It targets agent sandboxes, agent memory, and deployments on mobile and edge devices [1].

What is Relate?

Relate is described as “agent-native”: it was built around how agents actually work, so they spend their turns on the problem rather than fighting their tools [2]. Key characteristics include:

  • Lightweight graph storage: A graph can live in memory, a directory, or a single file. Developers can configure memory limits and remove the graph when a task finishes [2].
  • Structured results: Queries return JSON with capped row counts. Error messages explain how to correct a problem, and the database automatically detects agent harnesses, returning structured errors with explicit retry flags [1][2].
  • Shared graph updates: Transactions allow multiple agents to update the same graph without overwriting each other’s changes [2].
  • Coding-agent setup: A Relate skill can be installed into existing coding agents, helping them discover and work with the database [2].

Graph capabilities include multihop reasoning and GraphRAG (graph retrieval-augmented generation), which Neo4j says can help agents produce more accurate results and reduce hallucinations [1]. A single Cypher statement can combine vector similarity searches with graph traversals, and in-memory reads are sub-millisecond [1].

Relate is fully compliant with openCypher and more than 97% compliant with the Graph Query Language (GQL) [1].

Connectivity and environments

The database can run across Jupyter notebooks, web browsers (via WebAssembly), software containers, and edge devices. It includes a Model Context Protocol (MCP) server and a dedicated command-line interface for agent connectivity. Bolt and HTTP Query API v2 endpoints are also exposed, supporting integration with existing graph drivers and orchestration frameworks [1].

Hackathon and early demo

To coincide with the launch, Neo4j opened the Relate by Neo4j Hackathon 2026, a month-long online event starting 11 November at 6 PM Pacific on Devpost. The theme is “Context, Connected: agents that understand how your data connects.” Prizes include a Mac Studio and Mac mini devices [2].

A demo video published alongside the release shows Claude Code using a Relate skill to audit the 24 direct dependencies of Google’s Agent Development Kit (ADK). The agent spins up isolated, throwaway Relate graph databases for each dependency tree, checks for security advisories and unmaintained packages using data from deps.dev, reports the findings, and then cleans up the instances [3].

What it means for developers

For developers building AI agents, Relate removes the need to deploy and manage a separate database server. Because it runs as a library inside the application, graph lookups and vector searches become local operations rather than network calls or table joins [2].

The structured error handling and row limits are designed to keep agent interactions within context windows and help agents self-correct without custom parsing logic [1]. The ability to create ephemeral, single-file graphs suits sandboxed environments where an agent needs isolated storage for a session or a single task [2].

Relate is available now at relatedb.ai. The source code and installation instructions are public, and developers can try it directly from the terminal or by installing the skill into their coding agent [2].

Sources

  1. Neo4j launches Relate embedded database for AI agents
  2. Announcing the Relate by Neo4j Hackathon 2026: Build AI Agents on an Embedded Graph Database - Neo4j Graph Intelligence Platform
  3. Relate dependency analysis video
  • neo4j
  • relate
  • ai agents
  • graph database
  • embedded database

Researched and drafted with AI from the sources above.

Learn to build with AI

Hands-on tutorials on Cursor, Claude Code, AI agents and more.