Oct 3 — Oct 10, 2026
Activity this week was driven by the debut of the Lakebase serverless Postgres offering, substantial security hardening for Databricks Apps, and major API transitions introducing AgentKit while standardizing deployment runtimes.
1.Lakebase serverless Postgres debuts with agentic branching and Spark bulk loading
Databricks introduced Lakebase, a disaggregated serverless PostgreSQL engine targeted at operational and AI workloads. Updates highlight sub-second copy-on-write database branching for autonomous coding agents, Spark-parallelized bulk ingestion capable of loading terabytes into Postgres pages, and governed enterprise app deployment patterns.
Sources
- Lakebase and Agentic SDLC: Branching Databases for Coding AgentsNews · databricks-blog · Oct 8
- Load terabytes of data in minutes into Lakebase PostgresNews · databricks-blog · Oct 8
- How to build governed enterprise apps on Databricks with Replit and LakebaseNews · databricks-blog · Oct 8
- Building on Lakebase | Episode 1: Meet LakebaseVideo · Databricks Skill Builder · Oct 6
2.Genie expands with Meta Ads MCP integration and production ontology frameworks
Databricks expanded Genie's agentic ecosystem by adding Meta's Ads MCP server to Marketplace, enabling conversational marketing queries against governed warehouse tables. Case studies and architectural guides also demonstrated how Genie Ontology standardizes metric definitions and multi-agent enterprise workflows.
Sources
- How HelloFresh Optimizes 100% of Meta Ad Spend With Genie Agents on DatabricksVideo · Databricks · Oct 9
- How Genie Ontology Enhances AI Cowork | Databricks Genie One DemoVideo · Databricks · Oct 7
- Meta’s ads MCP server comes to Databricks: Put your customer intelligence to work in advertising campaignsNews · databricks-blog · Oct 6
- How Genie Ontology powers product development at DatabricksNews · databricks-blog · Oct 5
3.Databricks Apps adds GA On-Behalf-Of-User auth and Model Serving securables
Databricks Apps reached general availability for On-Behalf-Of-User (OBO) authorization, enabling applications to run actions strictly scoped to the caller's identity. In tandem, the official Python and Java SDKs added Model Serving endpoints as Unity Catalog securable types within app manifests, alongside deployment health check configurations.
4.AgentKit deprecates Mason across client SDKs as DABs drops legacy Terraform engine
Databricks updated its Go, Python, and Java SDKs to introduce the AgentKit workspace API while removing the deprecated Mason service in breaking releases. Simultaneously, Databricks CLI v1.20 eliminated the legacy Terraform execution engine, forcing asset bundle deployments onto the direct engine.