If you use AI coding agents regularly, you've likely watched your API token bills spike. Paying per-token just for an LLM to stumble through compiler errors and unit test fixes adds up fast.

That frustration led me to build Coding Sahayi—a native Windows C# IDE engineered specifically to save AI token costs.

The core idea: • Local-first execution: Keep heavy agent loops, AST analysis, and test-driven repairs entirely local via Ollama or LM Studio.

• Zero-token bug fixes: An interactive ConPTY terminal catches test failures and fixes them locally before saving, rather than burning cloud API credits on iterative syntax errors.

• In-IDE model adaptation: Save verified code solutions to SQLite and fine-tune your local model on your own hardware.

A special thanks to Alpamys Makazhan for open-sourcing Soup. His breakthrough work with layer streaming allows 8B models to be fine-tuned locally on 4 GB laptop GPUs without renting cloud hardware or paying cloud providers.

Building an entire desktop IDE from scratch has been an intense learning curve, but seeing the agent catch its own compiler bugs and fix them autonomously makes the late nights worth it. If you build developer tooling or work with local LLMs, I’d love to hear your thoughts on the architecture.

🌐 Project showcase: https://codingsahayi.github.io

💻 GitHub: https://github.com/codingsahayi/

🔗 Connect : https://linktr.ee/muhammedshabeer