Straight answers.
What CodAI is, what it runs on, and what it never sends anywhere.
General & licensing
- What is CodAI?
- CodAI (codaipro.com) is a free AI coding assistant that runs local GGUF models entirely on your own hardware — completion, chat, and code explanation with no cloud and no account.
- Is CodAI free for students and developers?
- Yes. CodAI is free under the GPL-3.0 license. There are no subscriptions, usage quotas, trial limits, or credit cards.
- How does CodAI work without internet?
- It runs lightweight models (Qwen2.5, Llama 3.2, DeepSeek-R1-Distill) on your CPU or GPU through a local llama.cpp runtime. Once a model is downloaded, every prompt is processed on your machine.
- Is my source code private when using CodAI?
- Yes. Zero telemetry or user data is transmitted — inference runs at 127.0.0.1 on your own hardware, and the GPL-3.0 source is available for you to verify that.
Models & privacy
- Which models can I run locally?
- Quantized GGUF models sized for laptop and desktop hardware: Qwen2.5-Coder-1.5B for code, Llama-3.2-1B for general chat, DeepSeek-R1-Distill-1.5B for reasoning, and Qwen2.5-0.5B for low-spec machines.
- Can I use CodAI inside my IDE?
- Yes. CodAI provides extension plugins and a local server API compatible with VS Code, JetBrains IDEs, and terminal workflows — all talking to the local runtime, never to a cloud.
- Do I need a GPU?
- No. The runtime is CPU-first; a GPU improves speed but is optional. The 0.5B model runs comfortably on machines with 2 GB of free RAM.
Installation & hardware
- How do I install CodAI on Windows, macOS, or Linux?
- Download the portable Windows build from the download page — one zip, no installer: unzip, drop a GGUF model into the models folder, double-click run.bat. macOS and Linux users can build from source. Full instructions are in the documentation.
- What hardware do I need for local models?
- Light models (0.5B–1.5B parameters) run in 2–4 GB of RAM on a modern dual-core CPU. Larger models (7B+) want 16 GB of RAM and benefit from a GPU, but are not required for CodAI’s core workflows.
Have more questions?