callstackincubator/cali

AI agent for building React Native apps

What it solves

Cali provides a role-oriented CLI for React Native and Expo teams to automate mobile quality assurance (QA), code reviews, performance audits, and feature implementation. It eliminates the manual effort of setting up deterministic environments for AI agents to test and review mobile applications across different platforms (iOS and Android).

How it works

Cali acts as an orchestrator that connects AI agents to specific tool packs and runtime contexts. It uses a shared cali-context.json file to define the workspace, repository, and mobile build details. Depending on the command used, Cali provides the agent with a bounded set of tools—such as agent-device for interacting with mobile devices or react-devtools for performance profiling—and generates structured reports (JSON and Markdown) that can be used locally or integrated into CI/CD pipelines like GitHub Actions and EAS.

Who it’s for

Mobile developers and QA engineers working with React Native and Expo who want to integrate AI-driven testing and review processes into their development workflow.

Highlights

  • Role-Based Commands: Dedicated entry points for qa, review, perf-review, and dev (implementation).
  • CI/CD Integration: Native support for GitHub Actions and EAS with automatic provider detection.
  • Deterministic Context: Uses a structured JSON contract to ensure agents have consistent information about the build and environment.
  • Tool Isolation: Assigns specific tool packs (e.g., repo-read, agent-device) to different roles to limit agent scope and increase reliability.
  • Multi-Platform Support: Handles both Android and iOS, including automatic platform inference from build artifacts.