Oz: The Orchestration Platform for Cloud Agents
Oz is the programmable orchestration platform designed for running and coordinating cloud coding agents at scale, enabling complex, multi-step automations.
What is Oz: The Orchestration Platform for Cloud Agents?
What is Oz: The Orchestration Platform for Cloud Agents?
Oz is Warp's advanced orchestration platform specifically engineered for managing and scaling cloud agents. It moves beyond simple command execution, providing a robust framework to spin up unlimited parallel, programmable, and fully steerable cloud coding agents. This platform allows developers to automate complex workflows that require coordination across multiple systems or repositories, transforming individual AI interactions into scalable, auditable business processes.
At its core, Oz bridges the gap between local development tasks and large-scale cloud automation. By leveraging 'Skills'—the building blocks of Oz agents—users can schedule, trigger, and manage sophisticated agent behaviors. With impressive adoption metrics, including over 350K agent conversations daily and a 97% acceptance rate for generated code diffs, Oz is proving to be a critical tool for modern software development teams looking to enhance productivity and consistency.
Key Features
- Agent Orchestration & Scheduling: Turn agents into automations by scheduling them to run like cron jobs, triggered via webhooks, or invoked through the API. This allows for reliable, recurring execution of complex agent tasks.
- Unified Control Plane: Start, monitor, and join agent sessions from anywhere—the Warp app, web interface, or mobile device. This centralized control ensures best-in-class agent steerability and session management.
- Multi-Repo Coordination: Agents built on Oz can work across multiple repositories simultaneously, enabling sweeping, coordinated changes across an entire codebase structure in a single operation.
- Fully Programmable Stack: Oz provides a comprehensive set of tools, including a dedicated CLI, SDK, and API, allowing developers to deeply integrate and build custom agent-powered applications.
- Multi-Model Compatibility: Oz is model-agnostic, supporting leading AI models like Claude, Codex, and Gemini, while also supporting industry standards like Skills for rapid onboarding and flexibility.
- Flexible Hosting: Choose to run your agents on Warp's infrastructure or deploy them securely within your own private infrastructure for maximum control and compliance.
- Auditability and Collaboration: All agent activities are auto-tracked, providing a clear audit trail. Furthermore, agents can be tagged directly within Slack messages, fostering seamless team collaboration.
How to Use Oz: The Orchestration Platform
Getting started with Oz involves defining your desired automation using Skills and then deploying and managing those agents through the unified control plane.
- Define Skills: Start by defining the core logic or task your agent needs to perform using Warp's 'Skills' framework. These Skills become the reusable components for your automations.
- Build or Configure Agents: Use the Oz CLI or SDK to assemble these Skills into a cohesive agent workflow. You can configure the agent to use specific AI models or access specific environments.
- Deployment and Triggering: Decide how the agent should run. You can manually initiate a session from the Warp app, set up a recurring schedule (like a cron job), or integrate it into CI/CD pipelines using webhooks or the API.
- Monitor and Steer: Once running, use the unified control plane (CLI, Web, Mobile) to monitor the agent's progress in real-time. You maintain full steerability, allowing you to intervene or adjust parameters mid-session if necessary.
- Review and Iterate: After completion, review the auto-tracked logs and audit trails. The high acceptance rate suggests that the outputs (like code diffs) are highly reliable, minimizing manual review time.
Use Cases
- Large-Scale Refactoring: Orchestrate an agent to traverse dozens of microservices or repositories, applying a standardized code change (e.g., updating an API client library version) consistently across the entire organization overnight.
- Automated Documentation Grooming: Schedule agents to run weekly, scanning newly merged pull requests, generating updated documentation, and creating corresponding draft updates for review, ensuring documentation stays current.
- Infrastructure Drift Remediation: Set up agents triggered by configuration file changes (via webhook) to audit deployed environments against desired state configurations, automatically generating remediation scripts or alerts.
- Onboarding and Setup Scripts: Create complex, multi-step setup agents that can provision new developer environments, clone necessary repos, install dependencies, and configure initial settings, all triggered by a single command or API call.
- Security Patch Deployment: Rapidly deploy security fixes by coordinating agents to identify vulnerable components across the stack and apply vendor-approved patches, followed by automated regression testing.
FAQ
Q: How does Oz differ from running a single AI prompt in a standard terminal? A: Oz provides orchestration. While a standard prompt runs a single interaction, Oz allows you to chain, schedule, monitor, and coordinate multiple agent runs across different environments or repositories, turning isolated tasks into reliable, scalable workflows.
Q: Can I use my own large language models (LLMs) with Oz? A: Yes, Oz is designed for flexibility. It natively supports major models like Claude and Gemini, but its architecture allows integration with various models, ensuring you can leverage the best tool for your specific automation needs.
Q: Is Oz secure for handling proprietary source code? A: Security is a priority. Oz offers flexible hosting options, meaning you can run agents entirely within your own infrastructure if required. Furthermore, all session activities are auto-tracked, providing a clear audit log of agent actions.
Q: What is a 'Skill' in the context of Oz agents? A: Skills are the foundational, reusable components that define an agent's capabilities. They allow you to modularize complex logic, making it easier to build, test, and deploy sophisticated agent automations based on established patterns.
Q: Can I start an agent session from my phone? A: Absolutely. Oz features a unified control plane that supports starting and tracking agent sessions from the Warp CLI, the web interface, and mobile devices, ensuring you have control wherever you are.
Alternatives
Dify
Unlock agentic workflow with Dify. Develop, deploy, and manage autonomous agents, RAG pipelines, and more for teams at any scale, effortlessly.
AakarDev AI
AakarDev AI is a powerful platform that simplifies the development of AI applications with seamless vector database integration, enabling rapid deployment and scalability.
Devin
Devin is an AI coding agent and software engineer that helps developers build better software faster.
PingPulse
PingPulse provides AI agent observability, allowing you to track agent handoffs, detect issues like stalls and loops, and receive alerts for misbehavior with minimal code integration.
SkillKit
SkillKit provides a universal set of skills allowing developers to write code instructions once and deploy them across 32 different AI coding agents, ensuring consistency and broad compatibility.
CodeSandbox
CodeSandbox is a cloud development platform that empowers developers to code, collaborate and ship projects of any size from any device in record time.