All posts

Does Lyzr’s control plane support bring-your-own API keys and locally hosted LLMs?

Does Lyzr's control plane support bring-your-own API keys and locally hosted LLMs?

Nirupam2 min read
Does Lyzr’s control plane support bring-your-own API keys and locally hosted LLMs?

Got a use case in mind?

8 weeks from use case to agents running in production.

Platform, people and FDEs, all in. Bring your environment. We'll co-build and stay until it's live.

Talk to Us

Free resource

101 AI Agents Use Cases

100+ production-ready AI agent use cases across Marketing, Sales, HR, Banking, Insurance, Operations and a lot more

Download now

Controlling your credentials and your model stack is a non-negotiable for enterprise security.

Yes, OpenController supports both.

You can bring your own API keys for any third-party service, and you can point agents to any locally hosted or private LLM.

This is a core design principle of OpenController by Lyzr, the control plane for enterprise AI agents. Every production agent stack needs a control plane.

OpenController is a framework-agnostic, cloud-agnostic one, built for enterprise governance that adapts to your infrastructure, not the other way around.

How OpenController Manages Your API Keys

You should never hardcode credentials into agent logic.

A request from an agent passes through the OpenController proxy, which pulls the API key from the vault and injects it before the call reaches the tool, so the agent never sees the credential.

OpenController separates secrets from code. It provides a secure vault for storing all your API keys, from OpenAI to a custom internal tool.

When an agent needs to call a tool, OpenController acts as a proxy.

It fetches the required key from the vault at runtime and injects it into the request.

The agent code itself never touches the raw credential.

This means you can rotate keys, manage access, and audit usage from a single, centralized place without ever redeploying your agents.

Connecting to Locally Hosted LLMs

OpenController is model-agnostic by design.

OpenController routes agent requests to both a public model API and locally hosted models like vLLM and Ollama in a private VPC, capturing all calls in a single trace.

It doesn't care if you're calling a public API or a private endpoint inside your own VPC.

As long as your locally hosted model exposes an API endpoint, OpenController can connect to it. This includes open-source models running on servers like vLLM or Ollama, or proprietary fine-tuned models deployed on private infrastructure.

You simply configure the endpoint and any necessary authentication within the control plane.

Lyzr's control plane then routes agent requests to your local model, just as it would to any external provider. All of these calls, even to local models, are captured in a single, end-to-end trace.

This gives you the cost and security benefits of local models without sacrificing governance and observability.

The Security and Compliance Advantage

Combining bring-your-own-key with local model support is critical for regulated environments.

A comparison showing sensitive data leaving the network to a third-party LLM versus staying inside the network boundary with a local model while OpenController records an audit trail.

It ensures that sensitive data, like PII or PHI, never has to leave your network boundary to be processed by a third-party LLM.

Your keys stay in your control. Your data stays in your control.

This architecture provides the technical foundation needed to enforce the strict access and data handling policies required for standards like HIPAA, GDPR, and SOC 2.

OpenController by Lyzr generates the audit trail that proves it.

Your Stack, Your Rules

Your infrastructure, your models, your keys: OpenController provides the governance layer without forcing you to change your stack.

It closes the gap between a working prototype and a production-safe agent.

See how it works. Talk to our team.

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

Got a use case in mind?

8 weeks from use case to agents running in production.

Platform, people and FDEs, all in. Bring your environment. We'll co-build and stay until it's live.

Talk to Us

Free resource

101 AI Agents Use Cases

100+ production-ready AI agent use cases across Marketing, Sales, HR, Banking, Insurance, Operations and a lot more

Download now

Continue reading

You might also like

Build with Lyzr

Try it in Agent Studio

From framework-agnostic design to production-grade agents, deployed in under 24 hours.