Lakebase
Product · SaaS / cloud software · Finance & accounting by Databricks
Serverless Postgres for apps and agents that scales automatically and branches like code.
A serverless, PostgreSQL-compatible database integrated with the Databricks lakehouse for AI agents and apps. It features instant branching, scale-to-zero compute, and unified transactional-analytical processing without ETL pipelines.
- Launched
- 2026-11-02
- Languages
- English
Category: Database
- Deployment
- Fully managed serverless on AWS, Azure, and Google Cloud
- Platforms
- API, Web (Databricks Workspace)
- Integrations
- Unity Catalog, pgAdmin, DBeaver, psql, Lakehouse Federation
- APIs and SDKs
- PostgREST API, Lakebase Data API
- Languages
- SQL
- Database Engine
- PostgreSQL-compatible
- Serverless Postgres engine with scale-to-zero compute
- Instant database branching for development and testing
- Persistent memory layer for AI agent chat sessions and state
Overview
Lakebase is Databricks' fully managed, serverless Postgres database designed for AI agents and applications. It unifies transactional and analytical data on the lakehouse architecture, allowing businesses to eliminate complex ETL pipelines by reading from a single governed copy of data. The platform features an elastic compute engine that scales to zero when idle and supports instant database branching for risk-free development and testing. Built for the AI era, Lakebase serves as a persistent memory layer for agents, storing chat sessions and context with low latency. It integrates natively with Unity Catalog for unified governance and supports standard Postgres extensions like pgvector and PostGIS.
Database Architecture
- Serverless Postgres Engine: Lakebase is a serverless Postgres engine that runs elastically on top of cloud object storage, allowing compute to dynamically adjust to workloads.
- Scale to Zero: Compute resources automatically scale to zero when idle, ensuring users only pay for what they actually use.
- Point-in-Time Recovery: Automated database snapshots enable precise point-in-time recovery, which can be validated in isolated branches before applying to production.
Developer Experience
- Instant Branching: Users can create instant database branches for development or testing without impacting live production workloads.
- Postgres Extensions Support: Lakebase supports standard Postgres extensions including pgvector for vector search and PostGIS for geographic data.
- Lakebase Data API: The PostgREST API allows for querying Lakebase directly with row-level security policies for fine-grained access control.
AI and Analytics Integration
- Agent State and Memory: Lakebase serves as a persistent state layer for AI agents, storing chat sessions, checkpoints, and messages with low latency.
- Unified Data Access (LTAP): The LTAP architecture allows transactions and analytics to read from one governed copy of data in open formats, eliminating the need for sync infrastructure.
- Zero ETL Pipelines: Transactional data is integrated directly into the lakehouse, removing the requirement for complex, custom ETL pipelines.
- Unity Catalog Governance: Lakebase integrates with Unity Catalog, allowing it to be registered as a catalog and used with Lakehouse Federation.
Use cases
- Developers use Lakebase as a persistent state layer to store AI agent chat sessions and context for durable across-session memory.
- Data teams eliminate ETL pipelines by using Lakebase to unify transactional and analytical workloads on a single copy of data.
- Software engineers create instant database branches to test application updates or agent experiments without impacting production data.
Specifications
- Deployment
- Fully managed serverless on AWS, Azure, and Google Cloud
- Platforms
- API, Web (Databricks Workspace)
- Integrations
- Unity Catalog, pgAdmin, DBeaver, psql, Lakehouse Federation
- APIs and SDKs
- PostgREST API, Lakebase Data API
- Sign-in and administration
- Integrated Databricks identity and authorization, row-level security policies
- Languages
- SQL
- Database Engine
- PostgreSQL-compatible
- Architecture
- Serverless with separated compute and storage
- Storage Format
- Open formats on cloud object storage
- API Support
- PostgREST API
- Governance Integration
- Unity Catalog
- Scaling Capability
- Scale to zero compute
- Recovery Type
- Point-in-time recovery with automated snapshots
How it works
- Provisioning and Setup: Users can start building with Lakebase by accessing it through the Databricks Data + AI Platform or starting a free trial.
- Development and Integration: Developers use familiar Postgres tools like psql or the integrated workspace editor to define schemas and manage data.
- Branching and Testing: Create instant branches for testing new features or agent logic without affecting the production database.
- Production Deployment: Deploy applications and AI agents that use Lakebase for low-latency state persistence and unified transactional-analytical processing.
Pricing
- Usage-based pricing
- Pay only for what you use
FAQ
What is Lakebase?
Lakebase is a database architecture that separates compute from storage, using a serverless Postgres engine on top of cloud object storage to enable instant scaling and branching.
Does Lakebase support Postgres extensions?
Lakebase supports standard Postgres extensions such as pgvector for vector search and PostGIS for geographic data.
Does Lakebase scale to zero?
Yes, Lakebase compute dynamically adjusts to workloads and scales to zero when idle, so you only pay for what you use.
Added by AntSpace from public sources (databricks.com).