Codex Pro User Shares Multi-Agent Setup: GPT-5.6 Main Agent With Three Specialized Subagents

pvncher · x · 2026-09-13

An OpenAI Pro plan user, following @pvncher's advice and using Astra to analyze usage patterns and benchmarks, shared a setup to better balance Codex usage: a main agent on GPT-5.6 Sol high (xhigh for some projects), plus specialized subagents—Luna max for general implementation, research, and running tests; Sol high for architecture/database/security work and reviewing Luna's output; and Astra high for planning, UI/frontend work, and complex long-running tasks. A practical recipe for quota-aware multi-agent orchestration.

Original post →

More from coding & agent

coding & agent channel →