What problem does it solve?
Autonomous loops addresses the complexity of running LLM-driven development workflows reliably and at scale by providing patterns and guardrails for iterative code generation, verification, and landing. It reduces manual coordination, prevents context bleed between stages, and automates recovery from CI failures and merge conflicts so teams can safely run agents over multiple iterations.
Core Features & Use Cases
- Sequential Pipeline: Chain isolated non-interactive steps for scripted CI-style automation and reproducible runs.
- NanoClaw REPL: Persistent session mode for interactive, stateful development using conversation-backed context.
- Infinite Agentic Loop: Parallelized, spec-driven generation with orchestrator-assigned uniqueness to avoid duplicate outputs.
- Continuous Claude PR Loop: End-to-end iterative PR creation, CI monitoring, auto-fix passes, and configurable stop conditions.
- Ralphinho / RFC-Driven DAG: Decompose RFCs into dependency DAGs, run tiered quality pipelines per work unit, and manage a merge queue with eviction and recovery.
- De-Sloppify Pattern: Dedicated cleanup pass to remove overly defensive tests and noise after implementation passes.
Quick Start
Use the autonomous-loops skill to set up a continuous Claude PR loop that creates iterative branches, runs tests, and auto-merges passing changes.