Summary
OpenAI's July 2026 Codex updates added end-to-end coding-task management inside ChatGPT for iOS (create, search, open, fork, and manage tasks, with diff filters, SSH host shortcuts, and per-task model settings) and, on July 9, extended Codex itself with indexed web search, configurable multi-agent delegation (disabled, explicit-request-only, or proactive per thread and turn), rollout token budgets that abort turns when exhausted, a revamped plugin browser, and usage-limit reset credits.
What changed
Beyond adding Codex task management to ChatGPT for iOS (create, search, open, fork, manage, plus diff filtering and per-task model settings), OpenAI's July 9 Codex update shipped indexed web search, per-thread and per-turn multi-agent delegation modes, configurable rollout token budgets with abort-on-exhaust, a reorganized plugin browser, and usage-limit reset credits.
Why it matters
Together these changes turn Codex into a governed, budgeted agent platform rather than an open-ended assistant: mobile task control lets developers steer async agents from any device, while rollout token budgets and delegation policies directly address runaway spend and unbounded multi-agent fan-out, the two hardest problems in production agent coding. Per-task model, reasoning, and credit controls show OpenAI treating agent runs as managed units of work.
Evidence excerpt
Codex adds usage-limit reset credits, a revamped plugin browser with smarter recommendations, rollout token budgets, multi-agent delegation controls, and indexed web search. App-server clients can configure multi-agent delegation as disabled, explicit-request-only, or proactive at the thread and turn level. Earlier, ChatGPT for iOS added support for creating, searching, opening, forking, and managing Codex tasks directly from a conversation.