</>MCP Agents Market
Plugin

lazycodex

by code-yeongyu3.4kTypeScriptUpdated 2026-08-09

The one and only agent harness for complex codebases. Project memory, planning, execution, and verified completion inside Codex.

Codex

LazyCodex is a Claude Code plugin that brings the OmO (oh-my-openagent) agent harness to Codex, providing project memory, planning, execution, and verified completion workflows for complex codebases. Inspired by LazyVim's approach to Neovim configuration, it offers a one-line installation that adds specialized skills, multi-agent orchestration, and intelligent model routing to Codex sessions. The plugin installs commands like $ulw-loop for verified task completion, $ulw-plan for strategic planning, and $start-work for executing plans with durable progress tracking. It includes hierarchical project memory through $init-deep, sub-agent roles (explorer, librarian, reviewer), and task-appropriate model selection to balance reasoning power with API quota efficiency.

Key Features

One-line installation via npx or experimental Codex marketplace integration
Multi-agent orchestration with specialized roles: Sisyphus, Hephaestus, Oracle, Librarian, and custom sub-agents
Intelligent multi-model routing across GPT-5.2, GPT-5.3-Codex, and GPT-5.4-mini based on task category
Verified completion loops ($ulw-loop) that run until Oracle confirms task success, capped at 500 iterations
Hierarchical project memory system ($init-deep) generating AGENTS.md context for complex codebases
Strategic planning workflow ($ulw-plan) and durable plan execution ($start-work) with Boulder progress tracking
Extensible skills library including LSP integration, AST-grep structural search, frontend polish, and code cleanup
Built-in diagnostics via npx lazycodex-ai doctor for installation health monitoring

Use Cases

  • 01Orchestrating complex refactoring tasks across large codebases with hierarchical project context
  • 02Planning and executing multi-step features with verified completion checkpoints
  • 03Running autonomous coding loops that iterate until evidence-based completion criteria are met
  • 04Balancing API quota usage by routing routine tasks to efficient models and complex reasoning to frontier models
  • 05Generating and maintaining project memory documentation that guides future agent edits
  • 06Spawning specialized sub-agents for exploration, library research, or code review within a single workflow

Related Plugins

View more

lazycodex — FAQ

What is LazyCodex?+

LazyCodex is a Claude Code plugin that packages the OmO agent harness for complex codebases, adding project memory, planning, verified completion loops, and multi-agent orchestration to Codex. It's inspired by LazyVim's approach to Neovim configuration, providing a curated distribution of skills and workflows for Codex.

How do I install LazyCodex?+

Run npx lazycodex-ai install in your terminal for the primary installation method. Alternatively, you can install experimentally from within Codex using /plugins, adding the marketplace https://github.com/code-yeongyu/lazycodex, then installing omo from the sisyphuslabs marketplace. Verify installation with npx lazycodex-ai doctor.

Which AI clients does LazyCodex work with?+

LazyCodex is specifically designed for Codex (Claude Code). It installs skills, hooks, and agent roles into the Codex environment and requires Codex's multi-agent tools to function.

Do I need API keys or special prerequisites?+

You need a working Codex installation with API access configured. LazyCodex leverages Codex's existing model access and does not require separate API keys, but the multi-model routing will use your Codex API quota across GPT-5.2, GPT-5.3-Codex, and GPT-5.4-mini.

Is LazyCodex free to use?+

Yes, LazyCodex itself is open-source and free under the MIT license. However, running the installed skills and commands will consume API credits through your Codex/OpenAI account based on the models selected by the task routing system.

How do I use LazyCodex commands after installation?+

Type $ in the Codex composer to browse installed skills. Commands are invoked with the $ prefix: $init-deep for project memory, $ulw-plan for planning, $start-work to execute a plan, and $ulw-loop for verified completion loops. Hooks announce themselves with LazyCodex status messages during sessions.

How do I install lazycodex?+

Open the source repository on GitHub and follow its README. lazycodex is a plugin — MCP Agents Market links you directly to the official repo.

Is lazycodex free?+

lazycodex is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.

Related searches