# Free local AI code assistant with Ollama

CodeSync SI's desktop app and CLI use Ollama by default. Scan, diagnose, safe repair, review, and generate without a paid model key.

## What is free

The desktop app and the codesync CLI do not require a CodeSync SI subscription. The default provider is Ollama at http://localhost:11434, or OLLAMA_BASE_URL if you set it. No OpenAI key is required for that path.

The recommended pull in this project is qwen2.5-coder:7b. qwen2.5-coder:1.5b is the smaller tag when memory is tight. The README lists approximate RAM and VRAM for those tags. This site does not claim a measured tokens-per-second number from a GPU we have not run.

## What the local pass does

scan walks the tree. diagnose reports the fixed findings. repair --apply writes the safe edits. review scores the snapshot. generate asks the model for files and, if the model is down or returns something unusable, writes the built-in template and labels it as a template.

Hosted plans are separate. They start at $30 per month and use Cloudflare Workers AI unless you store your own key.

## Related

- [/download](https://codesyncsi.com/download)
- [/docs](https://codesyncsi.com/docs)
- [Offline and private AI coding](https://codesyncsi.com/use/offline-private-ai-coding)
- [Ollama on a small machine](https://codesyncsi.com/blog/ollama-on-a-small-machine)