Tabby launched in October 2023 addressing a concern that had prevented many enterprises from adopting cloud-based coding assistants: sending proprietary source code to external APIs. Built in Rust with a self-contained Docker deployment requiring no external database, Tabby ran the entire inference pipeline within the organisation’s own network perimeter.
The tool provided IDE extensions for VS Code, JetBrains, and Vim alongside a team Answer Engine — a shared Q&A layer grounded in the team’s own code repositories. Repository-level RAG context gave completions and answers that reflected the team’s actual patterns and conventions rather than generic training data, addressing the customisation gap that had led some teams to prefer fine-tuned models over general-purpose completions.
Tabby reached 33,600 GitHub stars and Apache 2.0 licence, making it the leading open-source self-hosted coding assistant. Its success demonstrated a clear market segment: teams whose security posture, regulatory requirements, or code sensitivity made external API use unacceptable. The Pochi project (TabbyML/pochi), launched in 2025, extended Tabby with a full agentic loop capable of autonomous task execution in isolated git worktrees.