Enter your email address below and subscribe to our newsletter

Databricks open-sources Omnigent to unify AI agents

Share your love

  • Databricks open-sourced Omnigent, a meta-harness that sits above AI coding agents to let teams compose, control, and collaborate across them.x
  • Co-founder Matei Zaharia said it was built after observing engineers struggling to combine multiple agents into workflows at Databricks and Neon.linkedin
  • The alpha release, licensed under Apache 2.0, arrives just before Databricks’ Data + AI Summit opening Sunday in San Francisco.relvehq

Databricks Open-Sources Omnigent AI Agent Meta-Harness

Databricks released Omnigent on June 13, an open-source “meta-harness” designed to unify disparate AI coding agents under a single interoperable layer, addressing the growing challenge of agent fragmentation as engineering teams adopt multiple tools simultaneously.

A Layer Above the Agents

Omnigent sits above existing AI agents — including Claude Code, Codex, and Pi — providing what Databricks describes as a common interface for composing, controlling, and collaborating across them. The project, released under the Apache 2.0 license, allows developers to swap or combine harnesses without rewriting code, run agents locally or in sandboxed environments, and enforce cost budgets and data-centric policies at the meta-harness layer rather than through prompts.x

Databricks co-founder and CTO Matei Zaharia announced the release, writing that it was “based on the trend we saw with AI usage at Databricks and Neon: engineers were combining multiple agents into loops and workflows, and this was difficult above the harness layer”.linkedin

Composition, Control, and Collaboration

Omnigent’s three core pillars are composition, control, and collaboration. Composition enables multi-agent teams using different harnesses or models, switchable with one-line changes. Control introduces stateful policies and budget constraints enforced at the meta-harness level. Collaboration allows teammates to share a live agent session via URL with full history, enabling real-time review and steering from any device — terminal, web, desktop app, or phone.linkedin

The project is available in alpha and can be installed with a single command. A GitHub repository at omnigent-ai/omnigent and documentation site at omnigent.ai are now public.github

Timing and Context

The release comes just ahead of the Databricks Data + AI Summit 2026, scheduled for June 15–18 in San Francisco. It also follows Databricks’ broader push into enterprise agent infrastructure, including the general availability of Agent Bricks, its governed enterprise agent platform announced in April. While Agent Bricks focuses on enterprise governance, data context, and model routing within the Databricks platform, Omnigent occupies a different niche — an open-source tool for individual developers and teams working across multiple external agent harnesses.relvehq

Leave a Reply

Your email address will not be published. Required fields are marked *

Stay informed and not overwhelmed, subscribe now!