hezo-ai/hezo

What it solves

Hezo provides a secure, organized way to run teams of AI agents as if they were a corporate organization. It eliminates the need to manage multiple terminal tabs and manual coordination by providing a structured environment with org charts, project management, and budget controls.

How it works

Hezo operates as a server and web application that allows users to create projects and assemble teams of agents (such as a CEO, Captain, engineers, and designers). These agents run in sandboxed containers to ensure security. The system uses a "meta-harness" to standardize tools and memory across different AI models from providers like Anthropic, OpenAI, and Google.

Who it’s for

It is designed for users who want to deploy autonomous AI agent teams to handle complex tasks—such as app development or marketing—while maintaining strict control over security, spending, and approvals.

Highlights

• Organizational Structure: Uses a hierarchy of CEOs, Captains, and workers with customizable org charts.
• Security by Design: Employs an egress proxy and secret placeholders so agents never see real API keys or credentials.
• Sandboxed Execution: Every project runs in its own isolated container to prevent malicious or buggy code from affecting the host.
• Model Agnostic: Supports a wide range of providers (including local models via Ollama and LM Studio) through first-party CLI runtimes.
• Coded-in Control: Includes budget caps, approval gates for consequential actions, and a "Coach" that writes learned rules back to agents for continuous improvement.