Locode is an experimental CLI for AI assisted development. I am a builder and I built this while experimenting with a simple idea. Local models are fast and cheap. Claude is amazing for deep reasoning. What if a CLI could route simple tasks to a local model and only send complex problems to Claude? That is what Locode tries to do. The goal: • run simple coding tasks locally • route complex reasoning to Claude • reduce token cost and latency • keep developer workflows local first The project is still early and built mostly as a learning experiment. Fun fact. The entire project was built using Claude Code, which heavily inspired the workflow. Locode is also inspired by tools like Ruff, focusing on fast local developer tooling. There is a demo video in the README if you want to see it in action. Feedback, ideas, and contributors are very welcome. If you like the idea, please consider giving the repo a ⭐ to support the project.

Locode

Routes simple tasks to a local LLM, complex tasks to Claude

Routes simple tasks to a local LLM, complex tasks to Claude