lazycodex

[!NOTE] [OmO] 60K Stars: the terrifying token burner has arrived in LazyCodex.

Sisyphus Labs' OmO is the quality-obsessed agent harness whose public lore says it loved Anthropic models hard enough to get third-party clients blocked. Now that same OmO quality bar is available for Codex through LazyCodex.

If you wanted OmO but did not want the setup ceremony, start here:

npx lazycodex-ai install

Context: OmO 60K Stars on X

🚀 Install

One line. No global install, no npm i -g. Always use npx:

npx lazycodex-ai install

This is shorthand for npx --yes --package oh-my-openagent omo install --platform=codex. For a fully autonomous, no-TUI setup:

npx lazycodex-ai install --no-tui --codex-autonomous

Install from the Codex marketplace (experimental)

The npx installer above stays the primary path. As an additive, experimental alternative you can install from inside Codex itself: type /plugins, open the Add Marketplace tab ("Add a marketplace from a Git repo or local root."), and enter https://github.com/code-yeongyu/lazycodex, then install omo from the sisyphuslabs marketplace. Or from the CLI:

codex plugin marketplace add https://github.com/code-yeongyu/lazycodex
codex plugin add omo@sisyphuslabs

On the next launch, approve the omo hooks in Codex's startup review — hooks never run before approval. The first approved session prints LazyCodex bootstrap running in background — restart the session when it completes while a background worker finishes the setup (config blocks, agent roles, bin links, a pinned sg binary for the ast_grep MCP); restart when it is done. The marketplace path never touches Codex permission settings — autonomous mode remains the explicit npx lazycodex-ai install --no-tui --codex-autonomous choice.

Upgrade with codex plugin marketplace upgrade sisyphuslabs. The next startup review shows the hooks as Modified — expected after every upgrade — re-approve them and the following session re-runs bootstrap on the new version. If anything looks pending or degraded, npx lazycodex-ai doctor explains what and why. Full details: lazycodex.ai/docs.

Verify it worked

npx lazycodex-ai doctor

doctor prints the installation health report: plugin cache, hooks, MCP servers, agents, and config state. Inside Codex, type $ in the composer to browse every installed skill — init-deep, ulw-loop, ulw-plan, start-work, and the rest — and hooks announce themselves with LazyCodex(<version>): ... status messages during a session.

Uninstall

npx lazycodex-ai uninstall

Removes the installed plugin cache, bin links, agent roles, and the managed sections of ~/.codex/config.toml.

⚡ Commands

LazyCodex installs these as OmO commands for Codex. Invoke them with the $command syntax shown by the installer.

Full documentation lives at lazycodex.ai/docs.

Use the built-in workflows

LazyCodex should be judged by the features it actually installs. It is the Codex distribution for OmO's agent harness: project memory, planning, execution, verified completion, skills, hooks, model routing, and diagnostics.

1. $init-deep creates project memory

$init-deep generates hierarchical AGENTS.md context. It scores complex directories, writes local guidance near the code that needs it, and gives future agents landmarks before they edit. Type $init-deep in the Codex composer — the $ prefix is how every installed skill is invoked.

Use it when the repository is too large to explain from memory. Run it again when the shape of the codebase changes.

2. The three command pillars stay up front

Use $ulw-plan when the work needs decisions before implementation. It writes a plan to plans/<slug>.md and does not touch product code.

Use $start-work when a plan is ready. It executes the checklist with durable Boulder progress and stops only when the plan is complete.

Use $ulw-loop when the task should keep moving until the result is verified by evidence instead of a hopeful status update.

3. Skills cover specialized work

The command layer stays simple. The skill layer adds specialist judgment for the actual work:

4. Sub-agent roles ride Codex's native multi-agent tools

LazyCodex installs selectable agent roles into ~/.codex/agents/: explorer, librarian, plan, momus, metis, and codex-ultrawork-reviewer. Pick one by passing agent_type to Codex's spawn_agent tool — the child agent runs with that role's model and instructions:

spawn_agent({"message": "TASK: map the auth flow end to end.", "agent_type": "explorer"})

The installer exposes agent_type on multi_agent_v2 sessions (Codex hides it by default). If your Codex build's spawn tool has no agent_type parameter, describe the role inside message instead — the skills are written to fall back to that form automatically.

Start at https://lazycodex.ai.

💤 What is this?

LazyCodex packages OmO (oh-my-openagent) as the Codex agent harness for complex codebases.

Think LazyVim for lazy.nvim, but for Codex.

OmO is the agent harness: discipline agents, parallel orchestration, multi-model routing, skills, hooks, and verified completion. LazyCodex packages that harness for Codex.

"LazyVim made Neovim usable for the rest of us. LazyCodex does the same for Codex."

Credit: The LazyCodex name idea is inspired by LazyVim. The Ultragoal and UltraQA ideas are inspired by oh-my-codex, reimplemented from concept for this Codex setup.

🧩 What you get

🧠 Why different GPT models appear

Do not be surprised if an OmO/LazyCodex run shows models like gpt-5.2 with xhigh, gpt-5.4-mini, gpt-5.3-codex, or newer equivalents like gpt-5.5 with xhigh. That is intentional.

OmO does not blindly spend your best model on every subtask. Its source defines task categories and fallback chains so the agent can pick the most appropriate model for the job: quick routes to gpt-5.4-mini for small edits, ultrabrain uses a high-reasoning GPT model for hard logic, and agentic coding paths can use Codex-tuned GPT models when available. See openai-categories.ts and model-requirements.ts.

The point is quota discipline: use the strongest model when the task needs deep reasoning, use a cheaper/faster model when that is enough, and keep parallel agent work efficient instead of burning premium quota on routine steps. This is benchmark-driven routing, not random model churn:

• GPT-5.2 is documented by OpenAI as stronger at code review, bug finding, and complex tool use; the announcement notes that its maximum API reasoning effort uses xhigh.
• GPT-5.3-Codex is OpenAI's Codex-tuned model