FAQ
Short answers about LazyCodex install, OmO, ulw, hooks, doctor, tokens, Codex App, and OMX conflicts.
Start with the section that matches where you are stuck: Install and environment for setup, First use for choosing a command, Running and verification for behavior during a session.
Install and Environment
What is LazyCodex?+
LazyCodex is a Codex-focused light port that layers core OmO commands, skills, and hooks onto Codex. Think of it as a thin layer that helps Codex plan long work, keep going, and verify completion with evidence.
Is LazyCodex a replacement for OmO?+
Not exactly. It is closer to a light port of OmO for Codex. Full OmO has deeper orchestration; LazyCodex focuses on the parts that fit Codex directly.
Can I ask Codex to install it for me?+
Yes. Give Codex App or Codex CLI the LazyCodex GitHub link or lazycodex.ai and ask it to install LazyCodex. You can also run npx lazycodex-ai install yourself.
Do Codex App and Codex CLI install differently?+
Treat them as the same setup path. LazyCodex installs into the Codex environment; use Codex App or Codex CLI, whichever you normally work in.
Do I need Bun?+
No. You do not need Bun unless you are building LazyCodex from source. Normal installation and use follow the Node.js/npm npx path.
Does it work on Windows?+
It can, but native Windows is not the recommended path. If you must use Windows, use Codex and LazyCodex inside WSL2 Ubuntu. If install or runtime behavior gets strange, retry from WSL2 or a Git Bash-based shell before assuming LazyCodex itself is broken.
Which version should I look at?+
The docs show the plugin version. At the moment that is 4.12.1, taken from the LazyCodex OmO plugin package.
First Use
What kind of work fits LazyCodex best?+
It fits larger or longer work where planning, delegation, and verification matter. For small questions or one-line edits, ask Codex normally.
Should I use it brain-off or brain-on?+
Both work. LazyCodex is designed to run even when you delegate loosely, but it gets better when you can state the goal and completion criteria clearly. Do not study every internal skill first; learn the pieces only when you hit a real edge.
Which commands should I learn first?+
On a large repository, run $init-deep once first. Use $ulw-plan when the work needs decisions before code, $start-work to execute a written plan, and $ulw-loop when you want an open-ended task to keep going until evidence proves it is done.
How do I know the install worked?+
Open Codex, type $, and check that OmO commands and skills appear. In the CLI, typing ulw and seeing ultrawork mode enable is also a useful check. The first useful command is usually $init-deep.
What if commands do not appear in the $ menu after install?+
Open a fresh Codex session so the plugin reloads, then check whether omo hook approval is pending in the startup review. If commands still do not appear, run npx lazycodex-ai doctor to inspect install and skill loading state.
Running and Verification
Is npx lazycodex-ai doctor a real command?+
Yes. lazycodex-ai passes doctor through to the OmO doctor flow. It checks the local LazyCodex/Codex setup and can exit with an error when it finds problems.
Why do I need to approve hooks?+
Codex reviews hooks on startup. LazyCodex hooks do not run before approval, so approve the omo hooks after installation if you want the lifecycle checks to work.
What should I check after an update?+
If hooks show as Modified after an update, approve them again. If anything looks wrong, run npx lazycodex-ai doctor first to inspect install, hook, and skill loading state.
What if $ulw-loop finishes too shallowly?+
Repetition alone will not improve missing data or vague completion criteria. State the collection scope, required checks, and what to investigate when data is missing, or start with $ulw-plan first.
Do I need a higher plan or a huge token budget?+
Do not treat LazyCodex as a token-saving tool. It is closer to a workflow that spends a good model and enough tokens on planning, execution, and verification. For large work, split by work unit before a single thread gets too heavy.
How should I choose the thinking or reasoning level?+
Do not overthink it. Avoid low; use medium for normal work and high when mistakes are expensive or the task needs stronger review. Reserve xhigh for truly heavy work.
Does LazyCodex do computer-use?+
Yes, when the current Codex session has a computer-use tool available. LazyCodex can put that tool into the workflow, but browser or OS control still requires a callable computer-use capability.
Conflicts and Limits
Can I use it from Codex App, mobile, or remote sessions?+
LazyCodex installs inside Codex, so use it from whichever Codex surface you normally use. It pairs especially well with Codex App desktop/mobile remote flows, although a few capabilities may be desktop-app-only.
Team mode appears, but thread creation fails. What should I do?+
First update LazyCodex and Codex, then run npx lazycodex-ai doctor to check plugin, hook, and configuration state. Team mode can depend on Codex desktop app behavior, so do not assume the CLI and desktop app expose identical paths.
Can I use LazyCodex together with OMX?+
Not recommended. Keeping LazyCodex and OMX active together can conflict, spend tokens, and then fail. If the installer warns about another harness, clean one side up and run LazyCodex as the thin layer on top of Codex.