We are building the graph layer for AI-native software.
CrabGraph exists because the next generation of applications needs more than facts and embeddings. It needs the relationships between every entity, event, permission, and decision.
The best AI systems understand how data relates before they decide what to retrieve, explain, or automate.
Our mission is to make production graph systems ordinary.
Graph technology should not require a separate data island, a long migration, or a team of specialists. It should fit inside the stack teams already use.
Start where the code is
A graph database should be easy to add to a service, test locally, and ship incrementally.
Use the language teams know
SQL DDL, tables, foreign keys, and views are enough to describe a useful graph surface.
Serve many consumers
Agents, RAG systems, analytics, and APIs should reuse one governed graph layer.
What we believe
Our product decisions are shaped by the teams that have to operate graph systems after the demo works.
Graph should meet data where it lives
Modern data already sits in warehouses, lakehouses, and operational stores. CrabGraph brings traversal semantics to that data without demanding a copy-first architecture.
AI needs explainable relationships
When agents make decisions, teams need to inspect the path. Graph traversals give systems a concrete relationship trail instead of opaque similarity alone.
Developers should keep control
Start embedded, choose persistence, inspect schema, and move to cloud only when shared infrastructure is the right tradeoff.
Operations matter from day one
Performance, timeouts, metrics, auditability, and capacity planning are product features, not late-stage enterprise checkboxes.
Milestones
CrabGraph is early, focused, and built around a clear path from developer dependency to shared graph infrastructure.
Private beta
Validated embedded graph workflows with teams building recommendation, fraud, and AI retrieval systems.
Public beta
Added SQL views for derived relationships, JSONB property support, and stronger schema diagnostics.
CrabGraph 1.0
Released the embedded Gremlin-compatible server, persistent storage modes, and language SDKs.
Cloud graph layer
Bringing managed graph capacity to teams that need one shared relationship layer across many consumers.
Built by systems people for product teams.
CrabGraph is shaped by database internals, developer experience, and the daily reality of shipping AI features into production.
Storage
Durable graph state, WAL-backed persistence, and predictable local development.
Traversal
Gremlin-compatible APIs, query planning, and multi-hop performance.
Platform
Managed clusters, private networking, observability, and production controls.
Developer experience
Clear docs, fast setup, useful errors, and language-native package workflows.