Loop Engineering is a pattern library and CLI toolkit for designing systems that orchestrate AI coding agents around a codebase.
The tool addresses the problem of manually prompting AI agents by shifting focus to designing the loop itself—the system that discovers work, hands tasks to agents, verifies results, and persists state. Rather than typing the next prompt, developers define patterns that automate agent orchestration. The approach includes practical starters and CLI tools like loop-audit, loop-init, and loop-cost that help measure and optimize agent workflows. The tool supports multiple LLM providers including Claude, Grok, and others, with Claude as the default.
Loop Engineering suits teams managing repositories at scale who want to automate routine tasks like issue triage, dependency management, and PR handling without building custom orchestration from scratch. The pattern library covers specific jobs including keeping repositories healthy through daily triage, babysitting pull requests, and shipping features through refactor workflows. It is designed as a pattern library for operating agents around a codebase rather than as a general code rewriting tool, so it works best for teams with defined workflows they want to automate. The tool includes a thin GitHub Actions starter that requires minimal configuration files, making it accessible for teams wanting to start with low overhead.
The project maintains active development with regular updates to its pattern documentation and CLI tools. The codebase is written in TypeScript and includes comprehensive quickstart guides and job-specific documentation to help developers understand which patterns fit their use case. The tool provides a scoring mechanism through loop-audit that lets teams measure their agent loop design and track improvements over time.
