Choosing between Claude Code and Codex for a fairly large ongoing software project — need advice on workflows too
retained text
Hey everyone! I'm an individual developer working on a fairly large software project (enterprises level) that I've been developing for some time. I'm now considering subscribing to either Claude Pro ($20/month) for Claude Code or ChatGPT Plus ($20/month) for Codex. I'm having a hard time deciding which one would be a better long-term investment, so I'd appreciate advice from people who have actually used both. A little context about my project: \- Mobile application built with Flutter/Dart. \- Backend built using Java 21 and Spring Boot. \- Relational databases, with MySQL for development and PostgreSQL for production. \- Spring Security, JWT authentication, JPA/Hibernate and Flyway migrations. \- Fairly complex business logic involving state transitions, historical data integrity, concurrency, idempotency and database migrations. \- A reasonably extensive automated test suite (250+ backend tests). \- The project already has an established architecture, documented business rules and several completed development phases. This isn't a greenfield project where I just need an AI to generate boilerplate. A significant amount of implementation is already complete. The remaining work involves extending the existing architecture, handling complex edge cases, refactoring certain areas, improving synchronization and notification infrastructure, database migration work, testing and eventually preparing the application for production. I've never previously worked with such platforms i.e. Codex, Claude Code, Cursor, Antigravity. I'm now looking to settle on one primary coding agent, due to change in my personal commitments. I'd also appreciate advice on working strategies: \- How do you structure project instructions and documentation so the agent doesn't repeatedly misunderstand the architecture? \- Do you prefer giving agents one small task at a time or assigning larger features? \- How do you prevent agents from making unnecessary changes outside the requested scope? \- Do you use a separate AI for code review, or let the same agent implement and review? \- Are there any particularly useful workflows, configurations or habits that help reduce usage consumption? I'm working with a limited budget, so subscribing to both isn't really an option right now. I'd especially appreciate feedback from developers who have used both tools on real, established projects rather than just small demos or greenfield applications. If you had to pick only one at $20/month for this kind of work, which would you choose, and why? Also do not forget to drop strategies,. useful skills, and hacks, will mean a lot ☺️ Thanks!
document fetched 2026-11-10 00:24 · canonical https://reddit.com/r/developersIndia/comments/1xrld4e/choosing_between_claude_code_and_codex_for_a