Hezo – Self-hosted teams of AI agents that never see your real secrets logo

Hezo – Self-hosted teams of AI agents that never see your real secrets

Free
FreeFree tier
Inputs: text, codeOutputs: text, code
Type
Open Source

About Hezo – Self-hosted teams of AI agents that never see your real secrets

Hezo is a self-hosted platform for creating teams of AI agents that coordinate to complete tasks. It features a meta-harness with roles like CEO, captain, and worker agents that plan, edit, and review work. Agents never see real secrets due to a proxy system with AES-256-GCM encryption and Docker isolation. Users can set budgets per model, review work in real-time chat, and reuse skills. It works with providers like Anthropic, OpenAI, Google, and more. Setup is via a single curl command.

Key Features

Self-hosted with Docker isolation
Agents never see real secrets (proxy with placeholders)
Multi-agent team with CEO, captain, worker roles
Quality compounding through Coach agent that writes rules
Real-time chat and inline comment review
Live task board with full work history
Reusable skill files (built and borrowed)
Built-in MCP server and external connectors
Hard daily, weekly, and monthly budget caps per agent/project/model
Supports multiple model providers (Anthropic, OpenAI, Google, xAI, DeepSeek, etc.)

Pros & Cons

Pros
  • Data privacy and security: self-hosted, encrypted secrets, isolated containers
  • Agents never have access to actual API keys or secrets
  • Real-time oversight and collaboration via chat and task board
  • Quality improvement over time with automated Coach reviews
  • Flexible model provider support, easy to switch
  • Free and open-source (based on tool type)
  • Resource consumption tracking and budget controls
  • Extensible with custom skills and MCP tools
Cons
  • Requires technical expertise to set up and maintain self-hosted instance
  • No native support for local/self-hosted models yet (coming soon)
  • Dependence on external API keys and internet connectivity for model access
  • Limited to team structures defined by the system (CEO, captain, etc.)
  • No mobile app or GUI for non-technical users

Best For

Software development and engineering workflowsAutomating project management tasksCreating and reviewing code, docs, and launch postsCoordinating multi-agent teams for complex projectsSelf-hosted AI workforce for organizations with data privacy requirements

FAQ

Do I need to host my own models?
No — bring API keys or subscriptions for the providers you want. Hezo runs the agents; the models stay with them.
Can agents see my API keys?
No. Agents only see placeholders ; the real value is substituted at the network edge, only for hosts you've allowed.
How is Hezo secured?
Secrets are encrypted at rest (AES-256-GCM) behind a master key only you hold, every agent runs in an isolated Docker container, and all agent traffic passes through the egress proxy.
Is my data sent anywhere?
Hezo is self-hosted. Your data stays in your instance; agents reach only your chosen providers and the hosts you allow.
Can I run multiple projects?
Yes — each gets its own team and isolated container. Up and running in one command.