Tabby is an open-source, self-hosted AI coding assistant that gives teams Copilot-style completion and chat without sending code to a third-party cloud.
Tabby is an open-source coding assistant designed to run on your own infrastructure. It provides repository-aware code completion, a chat agent that can answer questions about your codebase, and IDE extensions for VS Code, JetBrains and Vim/Neovim, all backed by models you host yourself. Because it is self-hosted, source code never leaves your network, which makes Tabby a favorite in finance, government and other regulated environments.
Under the hood Tabby supports a range of open-weight models, offers enterprise features like SSO, audit logging and team analytics, and includes a repository indexing system so completions understand your project structure. It can be deployed with Docker or Kubernetes and scaled with GPU inference servers.
Tabby is designed to be transparent and controllable. Teams choose the underlying model, from compact completion models that run on modest hardware to larger models served on GPUs, and can fine-tune on their own repositories to improve suggestion quality. The platform includes completion and chat endpoints, repository indexing with a dedicated scheduler, and team analytics showing adoption and acceptance rates. Because there is no telemetry sent outside your network, security reviews are straightforward — a major reason Tabby is deployed in banks, government agencies and defense contractors.
Tabby is free and open source under Apache 2.0. Enterprise features such as SSO and advanced team management are available through paid enterprise plans.
Engineering organizations that want AI coding assistance but cannot or will not send their source code to external cloud providers.
Yes — the core product is open source and free to self-host; you only pay for your own compute and any enterprise add-ons.
Yes, Tabby exposes HTTP completion and chat endpoints that can be integrated into custom developer tooling.
Compared with cloud assistants like Copilot, Tabby trades convenience for full data control and model flexibility.
Tabby runs on CPU for small completion models, while larger models benefit from GPU inference; teams scale horizontally with dedicated inference servers as usage grows.