LazyCodex
The OmO agent harness packaged for Codex — project memory, planning, execution, and verified completion. LazyVim for Codex.
LazyCodex (LZX) packages OmO (oh-my-openagent) as the Codex agent harness for complex codebases. Think LazyVim for lazy.nvim, but for Codex.
OmO gives Codex a full agent harness: discipline agents, parallel execution, multi-model routing, a skills system, and hooks. LazyCodex packages that harness as a repeatable Codex setup you install with one npx line — no global install.
"LazyVim made Neovim usable for the rest of us. LazyCodex does the same for Codex."
What you get
| Feature | Description |
|---|---|
| Discipline agents | Sisyphus orchestrates Hephaestus, Oracle, and Librarian — a full AI dev team. |
| Parallel execution | Multiple agents work simultaneously on independent subtasks. |
| Multi-model routing | The right model per task category, chosen automatically. |
| Skills system | An extensible skill library for specialized work. |
| Hooks & lifecycle | Pre/post hooks for every agent action. |
| Zero config | Sensible defaults out of the box; override when you want. |
The command pillars
LazyCodex installs three command pillars plus a project-memory command for Codex. Invoke the pillars with the $command syntax the installer prints.
| Command | What it does |
|---|---|
/init-deep | Generates hierarchical AGENTS.md project memory. |
$ulw-plan | Prometheus strategic planner. Writes a plan to plans/<slug>.md; never writes product code. |
$start-work | Executes a plan until every checkbox is done, then prints ORCHESTRATION COMPLETE. |
$ulw-loop | Self-referential loop that runs until Oracle-verified completion. |
A typical flow: run /init-deep once for project memory, $ulw-plan "what to build" to decide the approach, then $start-work to execute it — or reach for $ulw-loop "task" when you want the work to keep moving until the result is verified by evidence.
Why different GPT models appear
OmO does not spend your strongest model on every subtask. It defines task categories and routes each one to the most appropriate model: quick lands on a small model like gpt-5.4-mini, ultrabrain uses a high-reasoning GPT model for hard logic, and agentic coding paths use Codex-tuned GPT models. The point is quota discipline — strong models for deep reasoning, faster models when that is enough.
Explore the docs
Getting Started
Install LazyCodex into Codex and run your first task.
Commands
The command pillars: ulw-plan, start-work, ulw-loop, init-deep.
Concepts
The ultrawork loop, discipline agents, and model routing.
Skills
review-work, remove-ai-slops, frontend-ui-ux, programming, and more.
Use cases & guide
Worked examples for everyday tasks.
Reference
Configuration and CLI.
Next steps
- Getting Started — install LazyCodex and run your first task.
- Hooks & lifecycle — when the lifecycle hooks fire.
- The underlying engine: oh-my-openagent (OmO)
- LazyCodex on GitHub
- Official site: lazycodex.ai. Maintained by Jobdori / Sisyphus Labs.