lazycodex
The one and only agent harness for complex codebases. Project memory, planning, execution, and verified completion inside Codex.
A one-command installer that adds a pre-configured multi-agent coding workflow to OpenCode.
LazyCodex is a one-command installer that sets up a pre-configured AI coding agent system inside OpenCode, an AI-assisted development tool. The name is a nod to LazyVim, a popular pre-built configuration for the Neovim editor: just as LazyVim removes the need to configure Neovim from scratch, LazyCodex removes the need to configure the underlying agent framework it ships with.
The core engine it installs is called oh-my-openagent (OmO), a separate project that coordinates multiple AI agents working in parallel. Once installed, LazyCodex adds three commands to your OpenCode session. The first, called ulw-loop, runs a task in a self-referential loop, checking its own output until it decides the task is done, up to 500 attempts. The second, ulw-plan, acts as a strategic planner and writes a structured plan file without touching any product code. The third, start-work, takes a plan and executes it step by step until every item is checked off.
Behind the scenes, OmO runs a team of named agents: Sisyphus orchestrates the others, Hephaestus handles code work, Oracle verifies results, and Librarian manages information. The system also supports automatic model selection, meaning it can route different types of tasks to different AI models without the user deciding.
The repository itself is a thin wrapper: the actual agent logic lives in oh-my-openagent, included as a code submodule. A Next.js website at lazycodex.ai hosts the documentation. The project is MIT licensed and is noted as being maintained by an AI assistant called Jobdori.
Where it fits
- Install a ready-made multi-agent coding setup in OpenCode with one command.
- Run a coding task in a self-checking loop until it reports completion.
- Generate a structured plan file before letting an agent touch product code.