Do agents have to run on Lyzr’s platform to be governed by the control plane, or can they stay on their own stack?

Adopting a new governance layer shouldn't force a complete rebuild of your agent stack.

No, your agents do not have to run on Lyzr's platform. They stay right where they are, on your own stack.

OpenController by Lyzr, the governance layer for production AI agents, is designed to integrate with your existing infrastructure, not replace it.

Every production agent stack needs a control plane. OpenController is a framework-agnostic, cloud-agnostic one, built for enterprise governance without vendor lock-in.

A Control Plane, Not a Hosting Platform

OpenController operates as a separate, neutral layer.

It does not host or execute your agent's code.

Instead, it integrates with your agents via a lightweight SDK. You wrap the agent's execution or its LLM calls with our SDK, which takes a few lines of code.

This wrapper acts as a checkpoint.

It intercepts requests, tool calls, and agent outputs before and after they happen.

This allows OpenController to apply your defined policies, log all activity for an audit trail, and enforce human-in-the-loop approval gates for consequential actions.

Your agent, built on LangChain, CrewAI, LangGraph, or a raw SDK, continues to run in your environment. OpenController simply gains the visibility and control needed to govern it.

Keep Your Stack, Gain Control

This decoupled architecture is a deliberate design choice.

It means you get governance without the pain of migration.

You avoid vendor lock-in. Your choice of agent framework, LLM, or cloud provider remains independent of your governance and observability layer.

Implementation is faster. You are not starting a migration project; you are adding a control layer to what you already have.

It provides unified governance across a distributed fleet. You can apply a single set of security and compliance policies to agents built with different tools and running in different parts of your infrastructure.

This gives you a single view to trace an agent's entire workflow, even when it calls other sub-agents.

Frequently asked questions

What does the integration actually look like?

It is typically a few lines of code.

You import the OpenController SDK and wrap the part of your code that executes the agent or makes the primary LLM call.

This is what connects your agent to the control plane.

Does this work with agents running in our own VPC?

Yes. OpenController is cloud-agnostic and built to govern agents wherever they run.

It connects to your agents in your VPC, on-prem, or on any major cloud provider.

How does this help with compliance?

By intercepting and logging every input, tool call, and output, OpenController generates the immutable audit trail that compliance teams need. This provides the evidence required to meet standards like HIPAA, GDPR, and SOC 2.

One Control Plane for Every Agent

The key takeaway: OpenController provides centralized governance without forcing you to centralize your agent architecture.

You keep the stack you've built. You gain the control you need.

See how it works with your stack. Let's talk.

Want to see how it fits your stack? Talk to us about OpenController by Lyzr.