Codex and ChatGPT Work share usage limits

OpenAI says Codex and ChatGPT Work usage draw from shared credits and limits, while API-key workflows continue to use standard API pricing. Why it matters: Coding-agent cost control is no longer just a per-tool setting. Shared credit pools mean model choice, context size, retrieval, caching, and tool use can affect the same budget across interactive and agentic work.

Try this: Before standardizing on a Codex workflow, document which work runs on subscription credits versus API keys, then set review points for large-context tasks and expensive model tiers.