In this video I show you how I stop swapping between Gemini, Claude (Anthropic), GPT, and even local models when I’m doing vibe coding in the terminal. The tool is LangCode (built with LangChain), and the whole idea is simple: bring multiple model providers into one CLI so I can stay inside my dev environment, chat with my repo, and let the tool handle the model selection instead of me constantly jumping between different CLIs. I walk through the setup (it’s basically a `pip install langchain-code`, then run `langcode`) and the key configs you actually care about: interaction modes (chat vs feature vs fix vs analyze), the “engine” choice (react for typical tasks vs deep for more complex multi-turn work), and router priorities (balanced, cost, speed, quality) when you want auto-routing. I also show where you put API keys in the local env file, how custom instructions work (MD file for project rules and style), and how MCP lets LangCode connect to tools like GitHub for PR workflows. Then I run a quick demo generating a `calculator.py` file end-to-end to prove the flow works from inside the terminal.