The State of Agentic iOS Engineering in 2026
Article Summary
Thomas Ricouard spent only 2% of his time in Xcode last year while shipping more iOS code than ever. His secret? Treating AI agents as tireless coworkers, not autocomplete tools.
In this comprehensive 2025 debrief, iOS developer Thomas Ricouard shares how he completely transformed his development workflow using AI agents like Codex and Claude Code. He covers everything from abandoning Xcode's text editor to building entire projects across multiple repositories with AI assistance.
Key Takeaways
- Xcode relegated to 2% of dev time, replaced by Cursor with AI agents
- XcodeBuildMCP enables 30+ minute unsupervised agent workflows with 90% success rate
- Codex 5.2 handles cross-repo edits, cloud-based bug fixes via iPhone
- Testing became essential: agents write tests faster than features
- Bootstrapped new Kotlin Multiplatform app with proper AI documentation from day one
AI-assisted iOS development has shifted from code completion to autonomous agents that build, test, and deploy features while developers focus on architecture and product decisions.
About This Article
Thomas Ricouard struggled with managing multiple open tasks and code reviews. He was constantly switching between separate backend, web, and mobile repositories, and keeping everything consistent across platforms was a real challenge.
He started using multi-repo code editing by referencing other repositories directly in his prompts. This let Codex review API changes overnight and add missing features across Kotlin, JavaScript, and Swift at the same time.
He no longer had to manually sync code across platforms. This meant he could work on multiple projects at once, and the straightforward implementation tasks that used to take hours now took much less time.
