The open-source agentic development environment, born out of the terminal. Build with agents, locally and in the cloud with Oz.
Intent
I need it when
Build and test agentic workflows across multiple LLM models and harnesses in a unified environment
Oz provides an open agentic development environment born from the terminal, supporting any coding agent, model, harness, repo, or tool. Developers can run agents locally or in the cloud, access frontier models (OpenAI, Anthropic, Google), and use custom inference endpoints, enabling flexible experimentation and production deployment.
Maintain data privacy and security while using cloud-based AI agents for development
Oz is SOC 2 compliant with Zero Data Retention agreements with LLM providers—they do not retain, store, or train models on customer data. Developers have full transparency via Network Log monitoring, and Enterprise customers get additional governance controls and the ability to bring their own LLM or self-host agents.
Scale AI-powered development across engineering teams with shared infrastructure and compliance controls
Oz Agent Platform offers team-level features including shared credit pools, admin-configurable data controls, SAML-based SSO, and Enterprise Analytics API. Organizations can enforce spend caps, manage multi-admin access, and self-host cloud agents on their infrastructure, supporting secure, compliant development at scale.
Reduce time spent on repetitive development tasks like debugging, deployment, and production diagnostics
Warp Terminal integrated with Oz agents enables developers to prompt their computer to build features, fix bugs, and diagnose production issues. Multi-agent orchestration and codebase indexing allow agents to understand context and execute complex workflows, turning tasks that normally take 10 minutes into seconds.
Accelerate software development by delegating coding tasks to AI agents while maintaining control and visibility
Oz orchestrates multiple coding agents (Claude Code, Codex, Warp Agent) in the cloud with centralized governance, usage visibility, and credit caps. Developers can join agent sessions with a single click and approve or deny agent actions, enabling faster feature development and bug fixes without sacrificing oversight.
Drop
Not a fit when
User requires a traditional GUI-based IDE instead of terminal-centric development environment
Organization needs on-premises-only deployment without cloud agent orchestration capabilities
Developer works exclusively on Windows and requires full feature parity with Mac/Linux versions
Team requires zero external LLM dependencies and cannot use OpenAI, Anthropic, or Google models
User needs a lightweight terminal replacement without AI agent integration or cloud infrastructure