localsetup

# Localsetup v2

<p align="center">
  <img src="assets/localsetup-v2-logo.png" alt="Localsetup v2" width="160">
</p>

<p align="center">
  <a href="LICENSE"><img src="https://img.shields.io/badge/license-MIT-blue.svg" alt="License"></a>
  <a href="https://agentskills.io/specification"><img src="https://img.shields.io/badge/Agent%20Skills-compatible-2ea44f" alt="Agent Skills compatible"></a>
  <a href="_localsetup/docs/PLATFORM_REGISTRY.md"><img src="https://img.shields.io/badge/platforms-cursor%20%7C%20claude--code%20%7C%20codex%20%7C%20openclaw%20%7C%20kilo%20%7C%20opencode-1f6feb" alt="Supported platforms"></a>
</p>

**Version:** 2.12.0  
**Last updated:** 2026-04-01

Agentic setups often share the same headaches: indeterministic outcomes, memory that compresses or decays, hallucinations, agents that drop context or ignore instructions, and difficulty scaling beyond a certain code size. Coordinating multiple agents so they follow patterns and run workflows reliably is harder still. Localsetup v2 targets these problems without adding much overhead.

The framework is repo-local: context, skills, and docs live in one folder in your project. Clone or move the repo and the setup moves with it. No home-directory state, no cloud dependency. Context is code, so you can audit what changed and tie specs and outcomes to git commits. It installs with one command and works the same across Cursor, Claude Code, OpenAI Codex CLI, OpenClaw, and OpenCode (add more via [one registry file](_localsetup/docs/PLATFORM_REGISTRY.md)). Safety and sandboxing are built in; when you import third-party skills, the framework runs security checks and heuristics before anything touches your agent. Tooling can be refactored or rewritten in Python and standardized even when sources disagree, and you can adapt it to your stack. The [public skill index](_localsetup/docs/PUBLIC_SKILL_INDEX.yaml) grows over time; you can add your own registry sources and combine or adapt skills as you like. One folder