the catalog · 🛠️ Dev tools

Tabby

★ 34k

Self-hosted AI coding assistant with completion, chat, and codebase context served from your own GPU

Tabby does the job of GitHub Copilot, self-hosted. It is Apache-2.0 (core)-licensed, so you can do essentially what you like with it, and setting it up is an afternoon, not a weekend. GitHub Copilot bills $10-19 for it. 34k stars on GitHub.

Open-source replacement for GitHub Copilot.

Licence
Apache-2.0 (core)
Self-hosting
Medium
The paid one costs
$10-19
GitHub stars
34k
view the code →official site →

Is it actually a replacement?

We have written a verdict on what it replaces — what you lose, and why people still pay:

How Tabby runs

Where it runs

Tabby runs on your server.

Installing it

Run the server from Docker or a standalone binary, then connect an editor extension.

Where your data lives

Models, configuration, logs, indexed source and Tabby's database live under TABBY_ROOT, which defaults to the user's .tabby directory.

What setup takes

Download and size suitable models, provide enough GPU or RAM, start the server, and configure each editor client.

The catch compared with GitHub Copilot

Completion quality and latency depend on the chosen model and hardware, and the GitHub and editor experience is less turnkey than Copilot.

Tabby questions

Is Tabby open source?

Yes — it is released under Apache-2.0 (core).

Can Tabby replace GitHub Copilot?

Our verdict on GitHub Copilot is YES. OpenCode (194k stars), Zed (88k stars), OpenHands (83k stars) are mature open-source replacements for GitHub Copilot — self-host in one command and stop paying $10/mo.

Can I self-host Tabby?

Yes — we rate it medium to self-host, and it runs on your server. Run the server from Docker or a standalone binary, then connect an editor extension.

More ai code editor / ide extension tools