Show HN: OneCLI – Vault for AI Agents in Rust

Introducing OneCLI: Securing the AI Agent Workflow

The rise of AI agents promises a new era of automation, where intelligent assistants can execute complex workflows, manage data, and interact with myriad APIs on our behalf. But this powerful new paradigm introduces a critical vulnerability: secrets management. How do you securely provide an AI agent with API keys, database passwords, and other sensitive credentials without hardcoding them into scripts or leaving them in insecure locations? This challenge is especially acute for developers and businesses building on modular platforms like Mewayz, where flexibility and security must go hand-in-hand. Today, we’re excited to share a solution we built in-house: OneCLI, a secure vault designed specifically for AI agents, written in Rust.

The Core Problem: Trust and Security in Autonomous Systems

When an AI agent needs to send an email via SendGrid, query a database, or update a project in a tool like Mewayz, it requires access to sensitive secrets. The traditional method of setting these as environment variables is brittle and insecure, especially when agents are scaled across different environments. Hardcoding is a non-starter. We needed a system that could centrally manage secrets, provide strict access control, and seamlessly integrate with agent execution flows. Our goal was to create a tool that acts as a trusted gatekeeper, ensuring that agents only receive the credentials they are explicitly permitted to use, and only at the moment they are needed.

"OneCLI is more than a keychain; it's a trust layer between your AI's intentions and its actions, ensuring security is a feature, not an afterthought."

Why We Built OneCLI in Rust

We chose Rust as the foundation for OneCLI for reasons critical to a security-focused application: performance, memory safety, and a robust ecosystem. AI agents operate in real-time, and any secret retrieval must be lightning-fast to avoid becoming a bottleneck. Rust’s zero-cost abstractions deliver the speed we need. More importantly, Rust’s compile-time memory safety guarantees help us prevent entire classes of vulnerabilities, such as buffer overflows, that could be exploited to leak sensitive data. This inherent safety is paramount when building the bedrock of your automation infrastructure. For a platform like Mewayz that emphasizes reliability, using a language engineered for safety and performance was the obvious choice.

Key Features of the OneCLI Vault

OneCLI is designed with a simple, Unix-philosophy approach: do one thing and do it well. It operates as a command-line interface that AI agents can call, returning the requested secret securely to stdout. Here’s what it offers:

• Centralized Secret Management: Store all API keys, tokens, and passwords in one encrypted vault, accessible via a simple CLI command.
• Scoped Access Tokens: Generate short-lived, permission-scoped tokens for individual agents, minimizing the risk of credential leakage.
• Audit Logging: Every secret access is logged, providing a clear trail of which agent accessed what secret and when, crucial for debugging and security audits.
• Seamless Mewayz Integration: OneCLI can be easily integrated into Mewayz modules, allowing agents operating within the business OS to securely retrieve credentials for internal or external services without any custom code.

Integrating OneCLI into Your Agentic Workflows

Using OneCLI is straightforward. An AI agent, whether it’s a Python script orchestrating a Mewayz workflow or a dedicated agent framework, simply makes a call to the OneCLI command. For example, an agent tasked with fetching data might execute onecli get database-password-prod. The CLI handles authentication and authorization, and if permitted, returns the secret directly to the agent’s process. This keeps secrets out of source code, environment variables, and agent memory until the very second they are required. This modular approach fits perfectly within the Mewayz philosophy, allowing you to compose secure, powerful business processes from discrete, reliable components.

We believe secure secret management is the missing piece for trustworthy AI agent deployment. OneCLI provides a robust, simple, and fast solution built with the safety of Rust. We’re open-sourcing OneCLI and are eager for the community to try it out, contribute, and help us build a more secure foundation for autonomous systems.

Frequently Asked Questions

Introducing OneCLI: Securing the AI Agent Workflow

The rise of AI agents promises a new era of automation, where intelligent assistants can execute complex workflows, manage data, and interact with myriad APIs on our behalf. But this powerful new paradigm introduces a critical vulnerability: secrets management. How do you securely provide an AI agent with API keys, database passwords, and other sensitive credentials without hardcoding them into scripts or leaving them in insecure locations? This challenge is especially acute for developers and businesses building on modular platforms like Mewayz, where flexibility and security must go hand-in-hand. Today, we’re excited to share a solution we built in-house: OneCLI, a secure vault designed specifically for AI agents, written in Rust.

The Core Problem: Trust and Security in Autonomous Systems

When an AI agent needs to send an email via SendGrid, query a database, or update a project in a tool like Mewayz, it requires access to sensitive secrets. The traditional method of setting these as environment variables is brittle and insecure, especially when agents are scaled across different environments. Hardcoding is a non-starter. We needed a system that could centrally manage secrets, provide strict access control, and seamlessly integrate with agent execution flows. Our goal was to create a tool that acts as a trusted gatekeeper, ensuring that agents only receive the credentials they are explicitly permitted to use, and only at the moment they are needed.

Why We Built OneCLI in Rust

We chose Rust as the foundation for OneCLI for reasons critical to a security-focused application: performance, memory safety, and a robust ecosystem. AI agents operate in real-time, and any secret retrieval must be lightning-fast to avoid becoming a bottleneck. Rust’s zero-cost abstractions deliver the speed we need. More importantly, Rust’s compile-time memory safety guarantees help us prevent entire classes of vulnerabilities, such as buffer overflows, that could be exploited to leak sensitive data. This inherent safety is paramount when building the bedrock of your automation infrastructure. For a platform like Mewayz that emphasizes reliability, using a language engineered for safety and performance was the obvious choice.

Key Features of the OneCLI Vault

OneCLI is designed with a simple, Unix-philosophy approach: do one thing and do it well. It operates as a command-line interface that AI agents can call, returning the requested secret securely to stdout. Here’s what it offers:

Integrating OneCLI into Your Agentic Workflows

Using OneCLI is straightforward. An AI agent, whether it’s a Python script orchestrating a Mewayz workflow or a dedicated agent framework, simply makes a call to the OneCLI command. For example, an agent tasked with fetching data might execute onecli get database-password-prod. The CLI handles authentication and authorization, and if permitted, returns the secret directly to the agent’s process. This keeps secrets out of source code, environment variables, and agent memory until the very second they are required. This modular approach fits perfectly within the Mewayz philosophy, allowing you to compose secure, powerful business processes from discrete, reliable components.