​Join us for Agentic + AI Observability meetup on Wednesday August 12th from 5pm - 8pm PST at the Databricks SF office,  an evening focused on agentic architectures and AI observability: architecture, governance, and real-world deployment of the open-source meta-harness for AI agents.

​This meetup is built for engineers, ML/AI practitioners, and AI startup founders who are already experimenting with agents (or planning to) and want to go deeper into the tech. We’ll cover real-world patterns, failure modes, and tooling for building a reliable meta-harness layer to manage agent execution, control and choice, state, and runtime policies across the broader ecosystem.

​Whether you’re at an early-stage startup or an established company, if you care about getting AI agents into production and keeping multi-agent systems healthy, this meetup is for you.

​Why you should attend

• ​See real architectures: Learn how teams are designing the meta-harness layer on top of data/feature platforms, isolated sandboxes, and tools, separating model reasoning from system execution.

• ​Learn how to observe what agents are doing: Go beyond logs and dashboards to structured traces, evals, and metrics that help you audit meta-harness behavior, enforce cost budgets, and track agent choices.

• ​Get hands-on with Omnigent: Watch live demos of Omnigent, Databricks’ newly open-sourced meta-harness for orchestrating cross-vendor agents and sandboxes.

• ​Connect with other builders: Meet engineers, founders, and practitioners working on similar problems, swap meta-harness design patterns, and find collaborators and potential hires.

​Agenda

• ​5:00pm: Registration/Mingling

• ​6:00pm: Welcome Remarks

• ​6:15pm: Talk #1: Meta Harnesses: The new layer for secure, collaborative, multi-agent coding (Corey Zumar, Staff Software Engineer at Databricks)

• ​6:45pm: Talk #2: Governing and Improving Production Agents with Omnigent and Codex (Wesley Pasfield, AI Deployment Engineer at OpenAI)

• ​7:15pm: Mingling with bites + dessert

• ​8:00pm: Night Ends

​Session Descriptions:

• ​Meta Harnesses: The new layer for secure, collaborative, multi-agent coding

  • ​Coding agents have become remarkably capable, and many software developers now depend on them for a large part of their day-to-day productivity. But the experience around them is still fragmented and hard to manage: for example, every harness and agent framework has a different interface, spending and security controls vary, sessions are often tied to a developer’s laptop, and teammates usually can’t see or steer them. In this talk, we demonstrate how Omnigent addresses these limitations by giving developers one place to use different coding agents and harnesses, build custom agents, run them remotely, collaborate on live sessions, and access them from any device. We’ll also share what we learned deploying Omnigent to thousands of engineers at Databricks.

• ​Governing and Improving Production Agents with Omnigent and Codex

  • ​Production agents need runtime governance and a way to improve from operating evidence. This talk shows an OpenAI Agents SDK agent running through Omnigent, which adds a shared session, tool policy, approval gates, and a durable decision record without replacing the agent’s reasoning or tools. MLflow captures the trace and reviewer feedback. Using a synthetic incident and simulated reviewer feedback, the demo follows a declined mitigation request into a tested agent update. Codex reviews the evidence, identifies a gap in the approval payload, updates the agent instructions and tool contract, and evaluates the new behavior. Omnigent governs the live action, the Agents SDK remains the agent runtime, and Codex drives the improvement loop. People retain authority over both the runtime decision and the release.