Codex

Install

curl -fsSL https://chatgpt.com/codex/install.sh | sh

Turn on the status line

/statusline

ref: https://www.jdhodges.com/blog/codex-usage-cli-status-line/

Setup Auto Compact

For example, gpt 5.6 ‘s effective context windows is “372,000-token context window with a 95% effective-context multiplier”, with a high cost threshold of 272,000. Long context exceed the limit will cause inaccurate response and low performance.

So it’s essential to setup auto compact for codex Edit ~/.codex/config.toml

model_context_window = 350000
model_auto_compact_token_limit = 300000

Ref: https://www.reddit.com/r/codex/comments/1uu7weh/possible_fixes_for_gpt_56_burning_through_your/