Tabby

Tabby is an open-source, self-hosted AI coding assistant designed to help developers write, understand, and improve code more efficiently. It serves as a privacy-focused alternative to cloud-based services like GitHub Copilot, allowing teams to run the entire assistant on their own infrastructure. The core functionality centers on real-time code completion: as a developer types, Tabby suggests inline code snippets, entire function bodies, or even multi-line blocks, based on the context of the current file and the project. These suggestions are generated by locally deployed language models, which can be fine-tuned or swapped out according to specific needs. Beyond completion, Tabby offers a conversational interface (often referred to as “Chat”) where developers can ask questions about their codebase, request explanations, generate documentation, or refactor existing code. It also supports an “Agent” mode, where the assistant can autonomously perform multi-step tasks — such as searching through files, making modifications, running commands, or even interacting with version control — under the developer’s supervision. This agent capability extends Tabby from a passive suggestion tool into an active coding partner. Tabby is fully self-contained: users download and run a server on their own machines or private servers, optionally with GPU acceleration for faster inference. The system supports multiple major language models, including those from the StarCoder, CodeGemma, and DeepSeek families, and it is compatible with both CPU and GPU environments (CUDA, Metal, or Vulkan). It integrates seamlessly with popular editors and IDEs through dedicated extensions, including VS Code, JetBrains IDEs, Neovim, and more. The assistant communicates with the IDE via a lightweight plugin th