codexray v0.1.5

codex-runner is now a haiku thin launcher (returns a job handle, never pipes Codex output). Main-session background run is the token-lean default: read only the final result; live streaming is free (node process, not a model).

codexray v0.1.4

Docs: how to set --model (incl. the fast gpt-5.3-codex-spark model), --effort, and --fast, with fast/cheap vs substantial presets — now in the README, the codex-run skill, and the site.

codexray v0.1.3

codex-runner is now a full drop-in subagent: one Agent dispatch runs the Codex task and returns the answer + real Codex token usage.

codexray v0.1.2

Fix: monitor manifest schema (top-level array + required description). Plugin now loads cleanly with the live-streaming monitor.

codexray v0.1.1

Fix: monitors.json top-level array so the plugin loads cleanly (streaming monitor). Install: /plugin marketplace add lidge-jun/codexray then /plugin install codexray

codexray v0.1.0

Observable Codex delegation for Claude Code: run Codex via codex exec --json with live progress + token visibility. Install: /plugin marketplace add lidge-jun/codexray then /plugin install codexray.