AbdoCode started with a simple question: what if AI coding agents could be genuinely useful without taking control away from developers? We were trying to solve a growing problem. Coding agents can read, edit, and run code, but developers still need to understand exactly what changed, what was executed, and how to safely undo it. We also wanted an experience that could work with local models, not only cloud providers. AbdoCode is a local coding agent for Windows, built with a small Rust kernel. It works with local models through Ollama or cloud models using your own API key. The agent can read, write, edit, and run code, while every action passes through a controlled gate with previews, validation, receipts, and rollback points. Our core principle is simple: The model proposes. The kernel decides. During development, our approach evolved from building a thin AI coding interface to designing a more structured execution system. We added adaptive safeguards, change reviews, browser capabilities, project templates, extensions, and integrations so developers can move quickly without giving up visibility or control. AbdoCode is built by TechnologyKSA, the Saudi technology company behind Mubarmij. The project source is published on GitHub for reading and local builds, with commercial use subject to the published licensing terms. We are launching AbdoCode on Product Hunt to meet developers who care about local-first AI, safer automation, and a more transparent relationship between AI agents and the code they change. We would love your feedback: What is the most important thing you need from an AI coding agent—speed, privacy, control, or reliability?

AbdoCode

ai

ai