ArangoDB
by Arango (formerly ArangoDB)
Native multi-model database now positioned as the context layer for enterprise AI
ArangoDB is a native multi-model database that stores documents, graphs, key-value pairs and vectors in one engine and queries them through a single language, AQL. It is aimed at teams who would otherwise run a graph database, a document store and a vector store side by side, and at architects building GraphRAG retrieval over enterprise data.
ArangoDB is a native multi-model database written in C++ that stores documents, graphs, key-value pairs, full-text search indexes and vector embeddings inside a single engine, queried through one declarative language called AQL. That combination is the product's original reason to exist: a team that would otherwise operate a graph database next to a document store next to a vector index can run one cluster and join across all of it in a single query, with native graph traversals rather than recursive SQL. The open-source core carries roughly 14,300 GitHub stars and 884 forks, and ships in Community and Enterprise editions, the latter adding replication, failover, hot backups, SmartGraphs, extended sharding and SLA-backed support. In October 2025 the company rebranded from ArangoDB to Arango and repositioned the database as the foundation of a broader Contextual Data Platform, with ArangoDB itself described as the contextual data foundation, the Platform Suite as the operations layer, and the Contextual Data Platform as the AI layer. Version 4.0, announced at NVIDIA GTC on 17 March 2026, added an Agentic AI Suite of more than twenty services including AutoGraph for automatic knowledge-graph construction from enterprise data, AutoRAG and Deep Search for routing a query to graph, vector or hybrid retrieval, and Arango Ada, a natural-language assistant that generates AQL. The company cites NVIDIA, HPE, the London Stock Exchange and the U.S. Air Force among its users, alongside more than 200 production deployments worldwide. Deployment spans self-managed Kubernetes, VMs and bare metal, the managed Arango Managed Platform, and OEM embedding for ISVs.
A data platform or ML architect who already needs graph traversal AND vector search over the same enterprise entities, and wants to stop operating two or three separate stores to get it.
One engine and one query language for graph, document and vector retrieval, which removes the synchronisation layer most GraphRAG stacks have to build and maintain between a graph database and a vector index.
At a Glance
- Category
- Data & Analytics
- Pricing
- Contact for pricing, Subscription, Free
- Target Market
- CTOs, Data Scientists, Enterprise Developers, Data Engineers, Solution Architects
- Deployment
- Self-hosted, Hybrid, Multi-cloud, Open-source
- Headquarters
- San Jose, United States
- Customers
- 200+ production deployments worldwide
Key Features
- ✓Native multi-model engine
Documents, graphs, key-value pairs, search indexes and vectors live in one storage engine, so a single query can join across all of them without a synchronisation pipeline.
- ✓AQL query language
One declarative language covers document filtering, native graph traversal and vector similarity, replacing the SQL-plus-Cypher-plus-vector-SDK split most stacks carry.
- ✓AutoGraph
Automatically constructs a knowledge graph from enterprise data at ingestion time, removing the hand-modelled ontology step that stalls most graph projects.
- ✓Deep Search and AutoRAG
Routes each incoming query to GraphRAG, VectorRAG or a hybrid path, so retrieval strategy is chosen per query rather than fixed at design time.
- ✓Arango Ada assistant
A natural-language interface that generates and optimises AQL and explores knowledge graphs, lowering the query-language barrier for business users.
- ✓SmartGraphs and enterprise sharding
Shards graph data so traversals stay local to a node in a cluster, which is what keeps multi-hop queries viable at scale in the Enterprise edition.
- ✓Flexible deployment
Runs self-managed on Kubernetes, VMs or bare metal, as the fully managed Arango Managed Platform, or embedded under an OEM agreement.
Capabilities
Use Cases
- •GraphRAG over enterprise knowledge
Ground an LLM in both the semantic content of documents and the relationships between the entities they mention, in one retrieval call.
- •Fraud and anti-money-laundering detection
Traverse transaction and counterparty networks several hops deep while filtering on document attributes in the same query.
- •Identity resolution and customer 360
Merge records scattered across CRM, billing and support systems into a single connected view of each customer entity.
- •Supply chain and dependency analysis
Model suppliers, parts and shipments as a graph to trace the blast radius of a disruption across multiple tiers.
- •Recommendation engines
Combine collaborative-filtering embeddings with explicit relationship data so recommendations can be explained by a traversable path.
Ideal For
Best For
- ✓GraphRAG and hybrid retrieval where an answer needs both semantic similarity and multi-hop relationship traversal
- ✓Fraud, AML and network-analysis workloads that are naturally graph-shaped but also carry large document payloads
- ✓Knowledge graphs built over heterogeneous enterprise data where the schema is still moving
- ✓Recommendation and identity-resolution systems that join entity relationships to embeddings
- ✓Consolidating an existing polyglot-persistence stack of graph plus document plus vector stores onto one cluster
Not Ideal For
- ✗Teams that require a permissive open-source licence — the core moved from Apache 2.0 to Business Source License 1.1, which forbids offering the database as a service to third parties until the four-year Apache conversion date
- ✗SaaS and DBaaS vendors who want to embed or resell the database, which the Additional Use Grant explicitly excludes without a commercial agreement
- ✗Organisations that only need a graph database and already have deep Cypher and Neo4j expertise, where AQL is a retraining cost with no offsetting benefit
- ✗Buyers who need published list pricing to budget — every tier routes to a sales quote
Integrations
Deployment
Market & Ratings
200+ production deployments worldwide
Market Analysis
Pros
- ✓One engine replaces a graph database plus a document store plus a vector index, removing the synchronisation code that usually sits between them
- ✓AQL handles traversal, filtering and vector search in a single language, so there is one thing to learn rather than three
- ✓Deployment is genuinely flexible — Kubernetes, bare metal, managed cloud or embedded — which suits regulated data that cannot move
- ✓Named in production at NVIDIA, HPE, the London Stock Exchange and the U.S. Air Force
Cons
- ✗The 2023 licence change is the dominant practitioner objection. One Hacker News commenter put it plainly: 'the license is awful and I don't feel like I can either open source or commercialize any of them until I'm running on an open source database.'
- ✗The same thread surfaces a trust problem rather than a technical one — 'Once a vendor has shown they have this attitude, I expect them to change their license for the worse in the future' — which is a hard objection to answer with features
- ✗No published pricing at any tier, so evaluation requires a sales conversation before a buyer can even size the spend
- ✗AQL is a proprietary query language with a far smaller talent pool and ecosystem than SQL or Cypher, making hiring and knowledge transfer harder
- ✗The rebrand to a 'Contextual Data Platform' layers several new product names (Platform Suite, Contextual Data Platform, Agentic AI Suite) over the database, and which capabilities sit in which tier is not clear without a sales call
Pricing
ArangoDB Community Edition
$0
- ✓Open source core under BUSL-1.1
- ✓Free for evaluation and non-commercial use
- ✓Internal production use permitted under the Additional Use Grant
- ✓No third-party database-as-a-service offering
ArangoDB Enterprise Edition
Contact for pricing
- ✓Replication and failover
- ✓Hot backups
- ✓SmartGraphs and fast traversals
- ✓Extended sharding and governance
- ✓SLA-backed support
Arango Platform Suite
Contact for pricing
- ✓Visualisation tools
- ✓High availability
- ✓Centralised orchestration
- ✓RBAC and SSO integration
- ✓Additional connectors and APIs
Arango Contextual Data Platform
Contact for pricing
- ✓GraphRAG and HybridRAG
- ✓Agentic AI Suite with 20+ services
- ✓AutoGraph and AutoRAG
- ✓Vector embeddings and GPU acceleration
- ✓MLOps and LLM integrations
No list pricing is published at any tier — every edition routes to a quote request. The Community Edition is free but is no longer Apache 2.0: since 3.12 the core is Business Source License 1.1, which permits internal production use but prohibits offering the database in a commercial service to third parties, and converts to Apache 2.0 four years after each release. Commercial redistribution or DBaaS use requires an Enterprise agreement, which is the practical cost driver for ISVs.
Security & Compliance
Connect
Sources
This page was written from 6 sources, 4 on domains other than arango.ai.
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Hex
Agentic analytics platform: AI notebooks for data teams and conversational Threads for everyone, grounded in governed semantic context
Unstructured
ETL for unstructured data: turn PDFs, tables, emails and scans into AI-ready JSON for RAG and agents
Tacnode Context Lake
PostgreSQL-compatible context lake giving AI agents shared, live, transactional state at decision time
Reducto
Agentic document parsing API that turns complex PDFs, scans and spreadsheets into citation-grounded data for AI, now with the 1¢-per-page r-1 model