How Warp's Oz Platform Just Changed Cloud Agents Forever
Most developers waste hours manually triggering agents and reviewing their output. Oz by Warp introduces autonomous cloud agents that generate weekly digests, validate PRs, and run scheduled workflows - all while you focus on building your product. Discover how this 3-file setup eliminates manual oversight and creates truly self-running automation.
The Agent Bottleneck Problem
Most developers using AI agents today face the same frustrating limitation - they're stuck manually triggering every agent run. Whether you're using Claude, Codex, or Cursor, you need to babysit your agents, giving them fresh instructions each time you want something done.
This creates an invisible productivity tax. Instead of focusing on building your product, you're constantly:
- Manually kicking off agent runs
- Reviewing intermediate outputs
- Copy-pasting results between sessions
- Remembering to run scheduled tasks
The hidden cost: Developers using traditional agents spend 30-40% of their time managing the agents rather than having the agents work for them. Oz eliminates this overhead by making agents truly autonomous.
What Makes Oz Different
While other platforms offer cloud agents, Oz introduces four revolutionary capabilities that change the game:
- Scheduled Execution: Agents can run on fixed schedules (weekly, daily, hourly) without manual triggers
- Event-Driven Workflows: Agents automatically respond to events like new PRs or database changes
- Agent Composition: Multiple agents can call each other, creating complex workflows
- Persistent Context: Agents maintain awareness of previous runs and system state
At the 4:30 mark in the video, you'll see how Oz automatically kicks off a PR validation agent after the digest generator completes - with zero manual intervention. This chaining of agents is where the real power lies.
Real-World Example: Weekly Digest Automation
The video demonstrates a production Oz workflow that:
- Queries a Convex database every week
- Generates a community digest with leaderboards and commentary
- Opens a PR with the formatted markdown
- Triggers a second agent to validate all links and build status
The result: A fully automated content pipeline that produces engaging, personality-filled digests (complete with inside jokes about the monthly theme) while catching technical issues before they reach production.
What's remarkable is how the LLM-powered agents add human-like touches to the output. At 7:45 in the video, you'll see how the digest includes witty commentary about a "passive aggressive task tracker that silently judges you" - something no template could produce.
The 3-File Magic Behind Oz Automation
The entire workflow shown requires just three files:
- weekly-digest.skill.md: Defines how to generate the community digest (under 200 lines)
- pr-validator.skill.md: Specifies PR validation rules and link checking
- oz-pr-validator.yml: GitHub Actions workflow that triggers the validator on PR open
These files leverage existing standards - the skill files use the same format as local agent tools like Cursor, while the workflow file is standard GitHub Actions. Oz's innovation is in how it orchestrates these pieces in the cloud.
At 12:20 in the tutorial, you'll see how the skill files contain simple markdown instructions that the agents follow autonomously, proving you don't need complex code to achieve powerful automation.
Agent Orchestration in Action
Oz's dashboard provides visibility into your agent ecosystem. You can see:
- Which agents are currently running
- Scheduled future runs
- Automatic triggers between agents
- Full execution traces for debugging
The video at 15:30 demonstrates how one agent automatically kicks off another - the digest generator triggers the PR validator without any manual intervention. This chaining is what transforms individual agents into a true automation system.
Key insight: Oz doesn't just run your agents in the cloud - it creates relationships between them that mirror how human teams hand off work.
Scheduling Made Simple
Traditional approaches to scheduling agents require setting up cron jobs or cloud functions. Oz simplifies this with:
- One-click scheduling interface
- Flexible intervals (15 minutes to weekly)
- Timezone-aware execution
- Model selection per schedule (Opus, Gemini, etc.)
At 18:50 in the video, you'll see how to set the digest agent to run every Monday at 9 AM UTC with just a few clicks - no infrastructure to manage.
Integration Possibilities
While the example shows GitHub and Vercel integration, Oz agents can connect to:
- Slack for notifications
- Linear for task management
- Any API-enabled service
The platform's strength lies in making these connections simple through its skill system. Rather than writing custom integration code, you describe what you want in markdown and let Oz handle the implementation.
Future-proof: Because Oz uses your existing agent skills, you can adopt new integrations as they're released without rewriting your automation logic.
Watch the Full Tutorial
See the complete walkthrough of setting up autonomous cloud agents with Oz. The video demonstrates key moments like agent chaining (4:30), skill file structure (12:20), and scheduling (18:50) that bring this automation to life.
Key Takeaways
Oz represents a fundamental shift in how we use AI agents - from manually triggered tools to autonomous team members that work while you sleep. The platform's ability to chain agents, respond to events, and maintain schedules unlocks automation that was previously impossible or required significant engineering effort.
In summary: With just three simple files, Oz lets you create self-running agent systems that generate content, validate work, and keep your projects moving forward - all without constant manual oversight.
Frequently Asked Questions
Common questions about Oz cloud agents
Oz stands out by enabling multiple agents to work together autonomously with scheduling, triggering, and composition capabilities. Unlike isolated agents from other platforms, Oz agents can call one another, be scheduled to run automatically, and integrate directly into your workflows without manual intervention.
Traditional cloud agents operate as single-use tools that require fresh instructions each time. Oz creates persistent agent systems that understand context and relationships between tasks.
- Agent chaining: One agent can automatically trigger another
- Event response: Agents react to GitHub PRs, database changes etc.
- Shared context: Agents maintain awareness of previous runs
The example workflow shown requires just three files: two skill definitions (one for generating weekly digests and one for PR validation) plus one GitHub Actions workflow file to trigger the agents automatically when PRs are opened.
These files use existing standards - the skill files match what you'd use with local agent tools, and the workflow file is standard GitHub Actions. Oz's innovation is in how it orchestrates these pieces in the cloud.
- weekly-digest.skill.md: Under 200 lines of markdown instructions
- pr-validator.skill.md: Defines validation rules and link checks
- oz-pr-validator.yml: Standard GitHub Actions workflow
Yes, Oz leverages existing agent skills from your code repository automatically. It uses the same skills definitions you'd use with other tools like Cursor or Codex, but orchestrates them in the cloud without requiring additional setup.
The platform detects skill files in your repo and makes them available in its interface. You can run these skills ad-hoc, on schedule, or as part of automated workflows - all while maintaining the same skill definitions across local and cloud environments.
- No lock-in: Skills work locally and in Oz
- Automatic discovery: Oz finds skills in your repo
- Consistent behavior: Same skills work everywhere
Oz agents can handle development tasks like PR reviews, data analysis, content generation, and workflow automation. They're particularly effective for repetitive tasks like weekly digests, PR validation, and scheduled reporting that would otherwise require manual intervention.
The platform excels at tasks that benefit from:
- Regular scheduling: Weekly reports, daily summaries
- Event response: PR validation, issue triage
- Multi-step workflows: Generate content → validate → deploy
- LLM enhancement: Adding commentary, analysis
Oz provides a simple scheduling interface where you can set agents to run at specific intervals (every 15 minutes, hourly, daily, weekly) without needing to configure cron jobs or cloud functions. The platform handles all the scheduling infrastructure automatically.
Key scheduling features include:
- Flexible intervals: From 15 minutes to monthly
- Timezone support: Run agents at local business hours
- Model selection: Choose different LLMs per schedule
- One-click setup: No infrastructure to manage
Yes, Oz agents can integrate with tools like Slack, Linear, and GitHub through their existing APIs. The platform also supports custom integrations through its skill definition system, allowing agents to work across your entire toolchain.
Common integration patterns include:
- Notifications: Post results to Slack channels
- Task management: Create Linear issues from agent output
- CI/CD: Trigger builds or deployments
- Custom APIs: Connect to internal tools
Oz supports multiple LLMs including Opus 4.6, Opus 4.6 Max (with 1M token context), GLM, Gemini, and Codex. You can select different models for different agents based on your needs for context length, reasoning ability, or coding specialization.
The platform lets you match models to tasks:
- Opus 4.6 Max: For long-context tasks like digest generation
- Codex: For code-focused validation tasks
- GLM/Gemini: For cost-effective high-volume tasks
- Per-task selection: Different models per agent
GrowwStacks specializes in implementing Oz workflows tailored to your specific business needs. We can design custom agent skills, set up automated workflows, and integrate Oz with your existing systems.
Our implementation process includes:
- Workflow analysis: Identifying automation opportunities
- Skill development: Creating custom agent skills
- Integration setup: Connecting Oz to your tools
- Ongoing support: Optimizing and expanding your agent ecosystem
Book a free consultation to discuss how autonomous agents can transform your operations.
Ready to Transform Your Workflows with Autonomous Agents?
Manual agent management steals hours from your week that could be spent building. Let GrowwStacks implement Oz workflows that generate content, validate work, and keep projects moving - all while you focus on what matters most.