Jul 22 — Aug 21, 2026
The period was dominated by the emergence of enterprise agentic infrastructure—highlighted by the open-sourcing of the Omnigent orchestration harness and enhanced Unity AI Gateway controls—alongside ecosystem-wide SDK upgrades to IAM v2 and local compute parity.
1.Databricks Open-Sources Omnigent Meta-Harness for Governed Multi-Agent Orchestration
Databricks introduced Omnigent, an open-source orchestration layer and runner for sandboxing, evaluating, and managing heterogeneous AI agents and Model Context Protocol (MCP) servers. The framework focuses on enterprise agent security by implementing contextual policies that detect and block data exfiltration when private internal context mixes with untrusted web content.
Sources
- Innocent until combined: Blocking the lethal trifecta with Omnigent Contextual PoliciesNews · databricks-blog · Aug 10
- Omnigent: Open-Source Meta-Harness for AI Agents | Matei ZahariaVideo · Databricks · Aug 10
- Building Agents on Databricks with Custom Apps and OmnigentVideo · Databricks · Aug 7
- Building Agents on Databricks with Custom Apps and OmnigentVideo · Databricks Skill Builder · Aug 6
- Claude vs GPT: Live AI Debate with OmnigentVideo · Databricks · Aug 2
2.Unity AI Gateway Adds Automated PII Guardrails, Smart Routing, and Strict Cost Quotas
Databricks expanded Unity AI Gateway capabilities with native input-level PII blocking to prevent sensitive information from reaching external model providers. Admins also gained granular spending thresholds with automated execution blocking, while Smart Routing was rolled out to direct model queries dynamically across the frontier price/performance curve for up to 30% cost reduction.
Sources
- Configure Budget Thresholds for AI ModelsVideo · Databricks Skill Builder · Aug 21
- Monitor LLM Usage & Prevent Data Leaks with TracesVideo · Databricks Skill Builder · Aug 19
- Databricks AI: PII Blocking for Data Privacy!Video · Databricks Skill Builder · Aug 15
- Smart Routing in Unity AI Gateway: Match frontier quality with 30%+ lower cost per taskNews · databricks-blog · Aug 13
- Don't Let AI Run Wild: Unity AI Gateway to the RescueVideo · Databricks Skill Builder · Aug 5
3.Databricks Genie Expands Beyond BI to Multimodal Agent Grounding and Pipeline Automation
Genie evolved from standard conversational BI into a versatile agent harness capable of reasoning across both structured Delta tables and unstructured documents stored in Unity Catalog Volumes. In parallel, teams are leveraging Genie Code to automate legacy PL/SQL and Power BI migrations into native Lakehouse pipelines and semantic models.
Sources
- Inbound Private Link now supports account-level Genie One, the account console, and custom URLsNews · databricks-blog · Aug 20
- Designing effective Genie Agents from a single promptNews · databricks-blog · Aug 19
- How a major freight railroad scaled pipeline creation with Genie CodeNews · databricks-blog · Aug 12
- How to ground Genie Agents in both structured data and documents without losing governanceNews · databricks-blog · Aug 10
- Power BI to AI BI: The Migration Shortcut You NeedVideo · Databricks Skill Builder · Aug 7
4.Tooling and SDKs Standardize on IAM v2 APIs, Git Deployments, and Environment Sync
Databricks SDKs across Python, Go, and Java, alongside the Terraform provider, rolled out unified IAM v2 endpoints with paginated listings for identity and workspace assignment management. Concurrently, the CLI and VS Code extension gained native capabilities to provision matching local virtual environments against remote runtime clusters.