Oz

Product Introduction

1. Definition: Oz is a cloud-native orchestration platform specifically designed for managing and scaling programmable AI coding agents. It falls under the technical categories of AI agent orchestration, cloud automation platforms, and developer productivity tools.
2. Core Value Proposition: Oz exists to enable developers and engineering teams to launch hundreds of parallel cloud agents rapidly, automating complex coding tasks (like generating PRs, fixing bugs, or auditing code) off local machines, thereby accelerating development cycles and reducing manual toil. Its primary keywords are cloud coding agents, agent orchestration, and AI-powered development automation.

Main Features

1. Programmable Agent Automations (Skills): Oz agents execute tasks defined in "Skills" – reusable, markdown-based instruction sets. Users can create custom Skills or use pre-built ones (e.g., accessibility audits, bug fixes). Agents run these Skills autonomously in the cloud. How it works: Users trigger Skills via CLI, API, web app, mobile, or scheduled events; Oz provisions the cloud environment, executes the Skill logic using configured AI models (Claude, Gemini, Codex), and reports results.
2. Unified Multi-Channel Control Plane: Users can start, track, join, and steer cloud agents from Warp Terminal, Oz web app, mobile device, or CLI. Real-time session joining allows direct intervention. How it works: Oz provides a centralized dashboard and APIs for agent lifecycle management across all interfaces, ensuring observability regardless of launch origin.
3. Flexible Infrastructure & Multi-Repo Operations: Oz supports agent execution on user-owned infrastructure (BYO) or Warp-managed cloud environments. Crucially, a single Oz agent can coordinate changes across multiple code repositories simultaneously. How it works: Infrastructure is defined via the Oz CLI/API. Multi-repo operations are handled by granting agents scoped access to relevant repositories, enabling atomic cross-repo changes within a single agent task.
4. SDK/API-First & Multi-Model Support: The platform offers a TypeScript SDK, CLI, and REST API for deep integration into custom workflows and building agent-powered applications. It is model-agnostic, supporting major LLMs and compatible with industry standards like OpenSkill. How it works: Developers use the SDK/API to programmatically create, trigger, and manage agents, integrating Oz logic into their own tools and choosing the optimal LLM per task.

Problems Solved

1. Pain Point: Scalability limitations of local AI coding tools. Running multiple complex AI agents locally drains resources and hinders parallel task execution. Oz scales agents off laptops onto cloud infrastructure.
2. Target Audience:
   • Software Engineers & Fullstack Developers: Automating PR generation, cross-repo refactoring, bug fixes.
   • DevOps/Platform Engineers: Building internal tools, scheduled monitoring agents (e.g., Sentry alerts), infrastructure-as-code checks.
   • Engineering Managers: Streamlining code reviews, documentation updates, and repetitive maintenance tasks across teams.
   • Data Scientists: Automating data pipeline monitoring, report generation (e.g., BigQuery insights to Slack).
3. Use Cases:
   • Automated PR Generation: Wake up to production-ready PRs created by cloud agents overnight.
   • Proactive Bug Fixing: Agents monitor error trackers (like Sentry) and automatically generate fixes.
   • Cross-Repository Refactoring: Implement a single change (e.g., API update) consistently across multiple services/repos atomically.
   • Scheduled Audits: Run daily accessibility checks or security linting using specialized Skills.
   • Slack-Integrated Workflows: Trigger agents directly from Slack messages for on-demand tasks.

Unique Advantages

1. Differentiation: Unlike basic AI coding assistants or CI/CD pipelines, Oz combines agent steerability, true multi-repo coordination, and cloud-native parallel execution into a single orchestration layer. It surpasses script-based automation by leveraging LLMs dynamically within auditable, programmable agents. Competitors lack this depth of orchestration and multi-channel control.
2. Key Innovation: The Skills framework provides a structured, reusable, and auditable way to define complex agent behaviors. Combined with the ability to join live agent sessions for intervention and the native multi-repo change capability, Oz offers a uniquely integrated and powerful approach to AI-driven development automation.

Frequently Asked Questions (FAQ)

1. How much does Oz cost? Warp has not publicly released detailed Oz pricing tiers. Enterprise plans and trials are available; contact sales via the "Book a demo" link for specific pricing based on usage (agents, compute time, features).
2. Is Oz secure for enterprise code? Yes, Oz is SOC 2 certified. It offers flexible hosting – run agents on your private infrastructure (BYO) for maximum control over data residency, network access, and secrets management, or use Warp's secure cloud.
3. What AI models does Oz support? Oz is multi-model compatible, supporting leading LLMs including Anthropic's Claude, OpenAI's Codex, and Google's Gemini out-of-the-box, allowing users to select the best model for each specific agent task or Skill.
4. Can I integrate Oz with my existing workflows? Absolutely. Oz provides a comprehensive TypeScript SDK, CLI, and API, enabling deep integration into custom tools, internal platforms, CI/CD pipelines (via webhooks), and communication tools like Slack for triggering agents.
5. How does Oz handle code changes across multiple repositories? Oz uniquely enables atomic multi-repo operations. A single agent can be programmed to access, modify, and commit/push coordinated changes across several designated repositories within one execution context, ensuring consistency for sweeping updates.