Arcade Data combines a multi-model database engine and an AI-native backend on a single open-source stack. Graphs, vectors, documents, workflows and agents — pre-integrated, so teams ship intelligent applications without assembling a stack of point services.
HACKBOX
From storage to AI orchestration, the components production AI applications actually need — pre-integrated and ready to run.
An AI-native backend that consolidates LLM orchestration, workflows, vector search, authentication, multi-tenancy and billing into one deployable system. Build production AI agents, copilots and pipelines in days, not quarters — without gluing together Supabase, n8n, LangChain and friends.
Visit arcadebrain.ai
The multi-model engine underneath: store graphs, documents, key-values, search indexes, vectors and time series in one ACID-compliant database. Queryable with SQL, OpenCypher, Gremlin, GraphQL or MongoDB-compatible APIs, at over 10 million records per second — with constant-time graph traversal.
Visit arcadedb.comThe workloads that need both connected data and reasoning — solved on one stack. Each of these uses ArcadeDB and ArcadeBrain together.
Vector-only RAG returns disconnected facts with no entity awareness. ArcadeDB stores your knowledge graph alongside its embeddings; ArcadeBrain orchestrates hybrid retrieval — graph traversal plus semantic search — before the LLM sees the context.
Result: grounded, explainable answers tied to verifiable structured data.
ArcadeDB's native graph engine catches multi-hop fraud rings in constant time. ArcadeBrain runs AI agents that triage flagged cases, query the graph for context, and write investigator-ready reports.
Result: faster investigation, lower false-positive cost, closed-loop detection.
Enterprise copilots need live operational data, not just stale docs. ArcadeDB stores records, vectors and the org graph; ArcadeBrain handles auth, multi-tenancy, LLM routing and tool use.
Result: copilots that answer "show me at-risk accounts in the EU pipeline" against real data.
Real-time recommenders need graph relationships, vector similarity and LLM reasoning simultaneously. ArcadeDB keeps all three in one store; ArcadeBrain orchestrates the pipeline so each recommendation carries an explanation trace.
Result: recommendations that reason, not just rank.
Extract structure from documents, resolve entities across sources, query with hybrid retrieval. ArcadeBrain workflows handle ingestion — extraction → embedding → entity resolution; ArcadeDB stores the resulting graph and vectors.
Result: every entity queryable both semantically and structurally.
Stop assembling fragile stacks of point services. ArcadeBrain provides the full agent backend — auth, workflows, vector search, billing — on top of ArcadeDB's operational store. One deployable system instead of seven.
Result: production agents shipped in days.
Graphs, documents, vectors, time series, key-values and full-text search in one ACID engine. No polyglot persistence, no ETL between stores.
Workflows, multi-tenancy, vector search, auth and billing — already wired together. Skip the assembly of Supabase + n8n + LangChain + Stripe.
ArcadeDB is Apache 2.0, perpetually free, with an explicit commitment never to relicense. Build with confidence on a stable, community-governed core.
10M+ records per second on ArcadeDB. Constant graph traversal. Embedded, self-hosted or managed — wherever your workload runs.
Evaluating ArcadeDB, deploying ArcadeBrain, or architecting a Graph + AI application that needs both — we'd like to hear from you.