# QodeX — the coding agent that stays on your side of the glass

QodeX is a local-first agentic coding CLI that defaults to running a local model through Ollama or LM Studio, edits and tests code with deterministic guardrails around the model, and can run its shell inside a Docker sandbox instead of directly on the host.

By default QodeX runs Qwen3-Coder through Ollama or LM Studio so a repository never becomes another vendor's training data, with Claude, GPT, Gemini and DeepSeek available only as opt-in cloud fallbacks. Named profile overlays switch between a fully local setup and a cloud-model-with-sandboxed-execution setup without becoming a second product, and a Docker-backed shell mode strips out the host's home directory, Docker socket, privileged mode and host networking so file edits stay inside the project.

A syntax gate, a completion gate and per-language auto-verification run around the agent's own loop so that even a weak local model cannot ship code that fails to parse or verify. A self-improvement loop captures the approach behind objectively successful tasks as quarantined skills, has an independent judge model promote the better ones through A/B comparison, and recalls similar past tasks through an episodic memory.

It is written in TypeScript, has over one hundred built-in tools, and is released under the Apache-2.0 license.

Website: https://qodexcli.github.io/QodeX/