有人用多个 AI 订阅撑过一整周高强度编程吗？Codex、Cursor、Claude Code 混搭方案讨论

I’m building an office portfolio management app for my business and have agents working on it for 12+ hours a day throughout the week. I’m trying to get the best quality work across that time without constantly hitting usage limits or paying for the highest tier everywhere.

My current setup is Astra for planning and specifications, then GPT-6 Sol High in Codex for most of the coding. When Codex usage gets low, I switch to Cursor with Grok 4.7 High. After bigger development batches, I use Claude Code on max for review, then send the worthwhile fixes back to Codex. So I pay for the $100 ChatGPT, and $20~ Cursor and Claude ones each.

They all use the same repo with shared plans and handover docs, so switching doesn’t mean explaining the whole project again. It works, but I’m wondering whether there’s a better way to divide the work and subscriptions.

Are others spreading heavy usage across a few lower-tier plans, or finding it better to pay more for one platform? I’m also interested in setups where a stronger model handles planning/manages the work and cheaper models like DeepSeek do more of the implementation.

I don’t mind paying for quality where it makes a difference. I’m mainly trying to avoid running out halfway through the week or saving money on a model only to spend hours fixing its work.

Would be good to hear from people doing similar hours: what’s your setup, roughly what does it cost, and what have you found actually works?