Can Lyzr’s Opencontroller platform call a custom ACI tool & keep the tool’s secret encrypted?

Connecting an AI agent to a custom tool often means hardcoding a secret.

Or worse, passing it in plaintext.

Yes, Lyzr's control plane can call any custom tool and keeps the tool’s secret encrypted within its own vault. The secret is never exposed to the agent, the LLM, or the transaction logs.

This is a core security function of OpenController by Lyzr, Lyzr's agent control plane. OpenController provides the governance layer between agent development and production, which includes secure credential management.

How Lyzr Secures Custom Tool Credentials

OpenController treats secrets as a managed resource, not as part of an agent's code.

The process is direct and secure.

1. You register your custom tool within OpenController.
2. The tool’s secret, like an API key or token, is stored in Lyzr’s encrypted vault. It is never in plaintext in your agent's code or configuration files.
3. Your agent is then configured to call this tool through the control plane, not directly.
4. When the agent needs the tool, it makes a request to OpenController.
5. OpenController intercepts the call, retrieves the credential from the vault, and injects it into the request to your custom tool.
6. The tool executes its function.
7. OpenController returns only the result to the agent.

The secret never leaves the secure boundary of the control plane.

It is never visible in the agent’s execution trace or the LLM's context window.

This Decouples Agent Logic from Security Risk

This model fundamentally changes how agents interact with sensitive systems.

Developers can build agents that use powerful internal tools without ever needing access to production credentials.

This separation allows you to keep the tools you already have and integrate them safely into your AI workflows.

Every tool call is logged for audit, but the credential itself is masked.

You have a complete record of which agent called which tool and when, without ever exposing the key.

This is critical for governing agent access to firewalled systems, like restricting access to sensitive finance data.

This method ensures your team can build complex agents that integrate with your firm's existing systems without creating new security vulnerabilities.

The Takeaway

A control plane’s job is to abstract away operational risk so your agents can function safely.

Securely managing credentials for custom tools is not an add-on.

It is a prerequisite for running enterprise-grade AI.

If you're evaluating how to connect agents to your proprietary systems, let's talk.

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