Hands-On with AgentControl
Agents break in production in all the new ways you'd expect—and a few you wouldn't. This workshop is about staying in control when they do.
In one hour, you'll get hands-on with AgentControl: Swap prompts and models without redeploying, set evaluation metrics that actually match what you're optimizing for, and watch the feedback loop close in real time.
Bring your skepticism. Leave with confidence.
About this workshop
In this hands-on workshop, attendees will master control over their agents and see how to change agent behavior in production without shipping code.
You'll learn how to:
- Create a config. Add AgentControl to your agent so its prompts, models, and tools can be managed from a UI in a few clicks.
- Update prompts, models, and tools without redeploying. Edit a variation's system prompt and watch the change take effect on the next chat call.
- Route users or segments to different models. Configure targeting so an enterprise cohort hits one model while free-tier users stay on another, no code change.
- Evaluate agents side by side. Run an offline evaluation against a labeled dataset and review row-level scores alongside aggregate results.
- Ship changes behind a guarded rollout. Configure a rollout with a model-as-judge guardrail, set a quality threshold, and watch automatic rollback trigger when the threshold isn't met.
We'll close with a quick tour of where to go next: agent-mode configs, agent graphs, custom judges, snippets, online evaluations, and the Agent Optimization private beta.
Prereqs: Just your computer. Accounts provisioned at check-in.
Speakers
What teams do in production with agents.
From release control to self-healing systems, these are the problems teams solve using runtime control—right from their existing tools.
Ship AI-built code with confidence.
Progressively release changes and roll back instantly based on real-time impact.
Control and govern AI agents in production.
Monitor agent behavior, detect drift, and take action without redeploying.
Experiment continuously.
Experiment with code and agents in production and optimize based on real-world results.
Optimize AI performance and cost.
Test prompts and models in production and dynamically route traffic to the best option.
Enable self-healing systems.
Automatically prevent issues, understand change behavior, and fix issues instantly with suggested changes and PRs.
