Govern AI Agents at Scale with Agent Registry
As AI agents spread across the enterprise, it gets harder to account for how many are running, what they cost, and whether they're safe. In this product video, we show how Agent Registry in the Fiddler AI Control Plane gives leaders one living inventory of every agent, with risk level, active alerts, and token usage side by side, so every agent stays accounted for and governed.
What you'll learn:
- A living agent inventory: See every agent application in Fiddler, with risk level, active alerts, and token usage in one view.
- Application-level context: See the projects, agents, and the models and tools each agent relies on.
- Operational and cost insights: Track traffic, token usage, and cost to see where spend is climbing.
- Quality and safety evaluations: Score agents for faithfulness, PII exposure, and safety with Fiddler Centor, Fiddler’s evaluation engine.
- Alert investigation: Zoom in on any alert to see captured, evaluated agent sessions and understand what happened without digging through traces.
Learn how the Fiddler AI Control Plane helps you govern AI agents at scale.
[00:00:00] AI agents are spreading across your organization. From first-party agents your teams build and deploy to third-party agents vendors bundle in and coding agents your developers use every day. Now, as adoption scales, you're probably finding it harder to answer questions such as the number of agents running in your environment, how much they cost, and whether they are safe.
[00:00:22] This observability gap leaves governance fragmented across teams and tools, and as a result, risky behavior can slip by, your costs can creep in, and compliance can become harder to prove.
[00:00:34] That's why we built Agent Registry in the Fiddler AI Control Plane. It's the first thing that you see when you log in.
[00:00:42] It's an up-to-date inventory of every agent running in your environment. And at a glance, you get to see all the applications running AI agents with their risk level, active alerts, and token usage side by side.
[00:00:57] Once you open an application, you'll get to see who owns it, which project it belongs to, the agents inside it, as well as the models and the tools those agents rely on. From here, you can also navigate to dashboards that offer crucial insights.
[00:01:08] For example, this operational dashboard shows metrics such as agent traffic, token usage, and costs. So if spending starts to climb, you'll see exactly where it's coming from.
[00:01:26] Similarly, this next quality and safety dashboard can evaluate agents for dimensions such as faithfulness, PII exposure, and safety, all scored by the Fiddler Centor Evaluation Engine.
[00:01:39] You can also zoom in on any alert to investigate the underlying cause. Every agent session is captured, and Fiddler performs evaluations on those sessions to help you understand what took place instead of you having to manually navigate through agent traces.
[00:02:00] This is how Agent Registry in the Fiddler AI Control Plane helps scale AI with confidence. Feel free to get in touch with us if you'd like to learn more.
[00:02:09] Thanks for watching.