0xwilliamortiz
Featured
openclaude-improved
OpenClaude is an open-source CLI coding agent that runs on any platform and supports a wide range of LLM providers, offering the same tools and workflows as Claude Code.
AI repository intelligence
FR-AI / ANALYSISWhy it is worth attention
It rigorously implements the insight that the agent's harness (rules, tools, mirror, guardrails) accounts for ~90% of the outcome—not the model—and was earned from months of real autonomous work, not theory.
Who it is for
- Developers using Claude Code, Codex, OpenClaw, or Cursor who want disciplined agent behavior
- Teams building AI-assisted software who need reliable handoff and verification workflows
- Tech leads or platform engineers integrating AI agents into multi-project environments
- DevOps and infrastructure engineers automating setup and maintenance with AI agents
Use cases
- Setting up a new project so the agent follows a consistent, rigorous development cycle (plan → implement → verify → ship)
- Running autonomous data crunching, marketing outreach, or document creation with safe failure handling
- Handing off complex work mid-session via a structured memory-first protocol for clean resumption
- Continuously improving agent behavior by editing the harness rules (fix the system, not the symptom)
Strengths
- Battle-tested over 6 months on real production work across multiple domains (software, devops, marketing, research)
- Modular design: a universal core plus 9 work-type profiles that adapt the rules to different kinds of work
- Philosophy of continuous improvement: each incident can sharpen a rule or add a gate, making failures less likely over time
- Cross-platform support: bash (Linux/macOS) and native PowerShell (Windows 7+) with identical flags and behavior
Considerations
- Requires an agent that reads CLAUDE.md/AGENTS.md (e.g., Claude Code) – not a standalone tool for human-driven development
- Learning curve to understand the harness philosophy and effectively edit rules rather than just fixing symptoms
- Safety mode configuration (cautious vs. trusted) demands careful per-machine decisions to avoid accidental data loss or exfiltration
README quick start
Quick start
Description
Universal, model-agnostic operating harness for AI agents (Claude, Codex, Gemini, …) — a lean core + work-type profiles assembled by one setup script.
Related repositories
Similar projects matched by category, topics, and programming language.
S40911120
Featured
recensa
Recensa is a self-hosted web viewer that indexes Claude Code session transcripts into a local SQLite database, enabling full-text search, replay, and audit of all past agent conversations without uploading data anywhere.
bjan
Featured
claudemux
Claudemux is a lightweight, dependency-minimal message bus that lets multiple Claude Code sessions communicate in real-time over tmux, enabling one session to ask another a question and receive the answer without manual intervention.