LazyCodex is a Codex agent harness for complex codebases. It is described as a light port of OmO’s Hephaestus, with an emphasis on goals rather than recipes and on verified completion. The documentation frames it as a way to run work in an outcome-first, evidence-driven manner, where a completion claim is not accepted until observable evidence supports it.
Its ultrawork mode is enabled by including ultrawork or the short form ulw in a prompt. In that mode, the harness switches to maximum-precision orchestration and follows a research, plan, execute, verify flow. The page shows Explorer and Librarian subagents gathering source truth, a Plan agent named Prometheus writing plans, and executor, TDD executor, QA executor, code review, and gate review roles taking part in the workflow. It also describes a self-referential loop that runs until verified completion and emits a DONE promise when a completion claim is reached, with an oracle required to verify it before the loop ends.
LazyCodex also provides a team mode for named teams of cooperating Codex threads. In that mode, the main session acts as the leader, splitting work, maintaining situational awareness, verifying and QAing results, relaying findings, and synthesizing the outcome. A bundled cross-platform script writes team state and an auto-generated member field manual. The documentation further mentions built-in skill coverage, automatic skill activation when tasks match a skill’s domain, and an explicit ulw-research mode for deeper research across codebase, web, official docs, and OSS repos with empirical verification and cited synthesis. It also says $init-deep generates hierarchical AGENTS.md context, $ulw-plan writes plans without product code, and $start-work executes a plan with durable progress tracking.
The project is presented as a docs site with a command-style interface centered on prompts and named workflows. It is linked to a CLI invocation using npx lazycodex-ai install, and the text includes a package install command that uses oh-my-openagent and omo install --platform=codex.
LazyCodex sits in PulseGate's Autonomous agents & workflows category. It focuses on coordinating autonomous agents for complex codebase tasks with memory, planning, and verification. LazyCodex is an open-source project aimed at developers working with AI agents and large codebases. The project is open source (MIT). LazyCodex is available on the web and the command line.
It is developed by code-yeongyu, and it first shipped in 2026. Development happens publicly on GitHub with 2.2k stars and 96 commits in the last 90 days. Key capabilities include agent orchestration, project memory, and parallel agents.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Closest matches by what these projects do