I Let My Users Build Their Own Features with AI Agents | Code with Beto

Here's how I connected my AI tattoo app to Warp’s new cloud coding agent platform, Oz, to automate feature development. When users submit feature requests, Oz spins up cloud agents that evaluate, implement, and open pull requests - all without me touching code manually.

If you want to scale your app’s development by letting AI handle feature requests, this walkthrough covers setting up Warp agents, configuring environments, and integrating the O SDK in a React Native TypeScript project. It’s for developers curious about cloud AI agents and automating coding workflows.

What's inside

How users unknowingly open pull requests with AI agents
Introduction to Warp’s cloud coding agent platform, Oz
Demo of the AI tattoo app and its feature request form
Setting up Warp cloud environments and connecting GitHub repos
Running and managing cloud agents from Warp’s UI and CLI
Using the O SDK in TypeScript to trigger agents on feature requests
Handling multiple parallel feature requests with autonomous agents
Real-time agent status tracking and usage monitoring

How users unknowingly open pull requests with AI agents

I built an AI tattoo app that generates $280 monthly recurring revenue, and now users can request new features that trigger AI agents to open pull requests automatically. The users don’t realize they’re actually contributing code changes because the AI agent evaluates the request, implements it, and submits the PR without manual intervention.

This automation removes the bottleneck of me manually coding every feature request, allowing the app to evolve faster while I focus on other priorities. It’s a practical example of how AI agents can empower users to build features autonomously.

Introduction to Warp’s cloud coding agent platform, Oz

Warp’s Oz platform runs AI coding agents in the cloud instead of locally. This solves common problems like noisy laptops, multiple terminal tabs, and losing track of agents. Oz lets you spin up many isolated cloud agents with full orchestration.

Agents can be triggered from the CLI, Slack, GitHub Actions, webhooks, or your own app code via the O SDK. This flexibility means you can integrate AI-driven coding workflows anywhere in your stack without local resource constraints.

Demo of the AI tattoo app and its feature request form

The AI tattoo app, Inkigo AI, lets users explore and try on tattoos. I added a “Request a feature” form where real users submit feature ideas. Currently, these requests send me Slack messages, and I implement them manually when I can.

I show selecting a tattoo, opening the playground, and submitting feature requests. This form is the trigger point for the AI agents to autonomously handle feature implementation going forward.

Setting up Warp cloud environments and connecting GitHub repos

Inside Warp, I create a new cloud environment tied to my AI tattoo GitHub repo. Warp supports selecting multiple repos if your backend and frontend live separately. The environment setup includes choosing a Docker image - Warp can suggest one automatically - and specifying setup commands like to prepare dependencies.

This environment runs the agent inside a container with full access to the codebase and GitHub, enabling it to clone, build, and push changes safely.

Running and managing cloud agents from Warp’s UI and CLI

Warp provides an agent management panel accessible via a shortcut (Shift+Command+M) where you can see running agents, their status, stop or cancel tasks, and track progress. You can also start agents from the CLI using commands like with prompts.

Agents run fully in the cloud, freeing your local machine. I show spinning up a test agent that responds with a greeting, demonstrating cloud execution and real-time status updates.

Using the O SDK in TypeScript to trigger agents on feature requests

In the app’s backend, I use the O SDK to instantiate a client with my Warp API key. When a feature request arrives, the SDK sends a prompt to the agent asking it to evaluate if the request is reasonable and, if so, implement it.

The SDK call includes environment ID, model ID (like ), and an optional base prompt describing the agent’s role as a helpful coding assistant. The response provides a job ID or state to track progress.

This minimal integration lets your app trigger autonomous coding agents programmatically on demand.

Handling multiple parallel feature requests with autonomous agents

I demonstrate submitting two different feature requests simultaneously from separate devices. Warp spins up two parallel cloud agents working independently on each feature.

This concurrency means your app can scale feature development without waiting for one request to finish before starting another. You can monitor all agents’ progress and usage centrally in Warp.

Real-time agent status tracking and usage monitoring

Warp’s UI and CLI let you track each agent’s state, cancel tasks, and view logs in real time. You can also monitor AI credit usage and other metrics to manage costs.

This visibility is crucial for maintaining control over autonomous agents running in production and ensuring they deliver value without surprises.

Resources

CourseReact Native course

Fundamentals through shipping: the concepts behind the prompts, with lifetime access.

Premium resourcePro Membership

Get early access to AI automation tutorials and exclusive coding workflows.

Let's connect!