Tabby

Tabby is an open-source AI coding assistant built by TabbyML, Inc., designed to bring AI-powered development capabilities to teams that need transparency, security, and deployment flexibility. It runs self-hosted or in the cloud, supports consumer-grade GPUs, and requires no external database or cloud service dependencies. The project has accumulated over 33,500 GitHub stars, signaling strong community adoption.

What It Is

Tabby is a self-hosted alternative to proprietary AI coding assistants. It provides real-time code completion, an in-IDE answer engine, inline chat, and data connectors—all within a single deployable server. Because the core platform is open source under an OSI-approved license, teams can audit the software supply chain, customize behavior, and run it entirely on their own infrastructure. TabbyML positions Tabby as a "secure, flexible, and transparent" option compared to closed-source tools.

Core Feature Set

Tabby ships four primary capabilities accessible from supported IDEs:

• Code Completion — Predicts the developer's next move in real time, adapting to coding style and project context.
• Answer Engine — Answers coding questions inline without requiring the developer to leave the editor; it reads repository files to ground responses in actual project code.
• Inline Chat — Enables real-time AI-assisted conversation tied directly to code context inside the editor.
• Data Connectors (Context Providers) — Connects to external sources such as GitHub, GitLab, and Git repositories so the assistant understands project-specific documentation and configuration.

A new Pochi agent feature, announced July 18, 2025, extends Tabby into autonomous task execution—planning, executing, and checking in on tasks like a human teammate, integrated into existing tools and workflows.

IDE and Language Support

Tabby integrates with a broad range of editors and IDEs:

• VS Code, Neovim
• IntelliJ-family IDEs: IntelliJ IDEA, PyCharm, WebStorm, GoLand, PhpStorm, CLion, Rider, RubyMine, AppCode
• Eclipse, Android Studio

The tool supports the languages, libraries, and frameworks developers already use, with no lock-in to a specific language ecosystem.

Deployment Model

Tabby is designed for local-first and on-premises deployment. Key deployment characteristics include:

• No external DBMS or cloud service required
• Runs on consumer-grade GPUs
• Integrates with Cloud IDEs
• Self-onboarding via Docker (documented quick-start path)
• Cloud-hosted option also available through Tabby Cloud, which uses usage-based billing tied to LLM token costs

The self-hosted path gives security-conscious teams full control over data residency and model selection. TabbyML partners with Mistral AI, DeepSeek, and Qwen for model access.

Update: Pochi Agent Launch (July 2025)

On July 18, 2025, TabbyML launched Pochi, described as a "full-stack AI teammate" that takes on tasks autonomously—planning, executing, and checking in—fitting into existing tools and workflows without requiring new tooling. Pochi is available on Tabby Cloud and billed on a usage-based model tied to LLM token consumption, with $20 in free monthly credits included. Tab Completion remains free with no usage limits on the cloud plan.

Open-Source Lineage and Community

Tabby's core server is publicly available on GitHub under the TabbyML organization. The project encourages community participation through a Slack workspace, a blog, and open GitHub contributions. The homepage invites developers to connect with other users, get answers, and collaborate on open-source projects. TabbyML also publishes a newsletter for product updates.