tabby

Tabby positions itself as an on premises alternative to GitHub Copilot. It runs self contained, speaks an OpenAPI interface, and the README trails a long series of release updates.

A self hosted AI coding assistant. It is open source, runs on premises, and the main claims come in three: self contained with no database or cloud service required, an OpenAPI interface that slots into existing infrastructure such as a cloud IDE, and support for consumer grade GPUs. The primary language is Rust.

The news section reads like a release log. Recent entries connect GitHub issues to tasks so pull requests can be created straight from the sidebar with CI, lint, and test results broken down, let GitLab merge requests be indexed as context, accept documentation through REST APIs, and turn Answer Engine messages into persistent, shareable pages. Earlier versions added LDAP authentication, a notification box, and llamafile deployment support.

The documentation covers installation, IDE and editor extensions, and configuration, and there is a promise that you can run it in about a minute, with inference type and parallelism as documented options. For anyone who cloned the repository, the README suggests a git submodule update to pull in all submodules before building.

Twitter or X for general engagement, LinkedIn for updates, and a newsletter. The project is developed by TabbyML, and the whole thing is framed as an alternative for teams that want AI assisted coding without handing their data to a third party.

The steady release cadence and the repeated focus on context, GitLab and GitHub, and answer persistence stand out. For teams that want coding help on their own machines, the README makes a coherent case.

• Source 1 · Project repository
• Source 2 · README