Why build this when we have Cursor, Windsurf, Claude Code, etc.? Those are great, but sometimes you just want your existing SOTA LLM to handle multi-step tasks without switching tools. All you need is LLM.
Example: ``` llm loop "Create a Flask web application" ```
The AI will create files, write templates, generate requirements.txt, and provide run instructions - all in one command.
Key features: - Chains multiple tool calls within conversations - Built-in safety with approval mechanisms - Works with existing LLM CLI tools and custom functions - Use any model as an autonomous agent
I use it for project setup, code generation, and analysis tasks where you'd normally need multiple back-and-forth prompts or a separate AI coding tool.
GitHub: https://github.com/nibzard/llm-loop PyPI: https://pypi.org/project/llm-loop-plugin/
Thoughts on just LLM vs. dedicated coding agent?