The open-source alternatives to GitHub Copilot are Tabby, twinny, Kilo Code, Cline and Zed. For Copilot's core, inline completion in the editor, twinny does it in VS Code from a local model under MIT, Tabby serves it to VS Code, JetBrains and Vim from a self-hosted server, and Kilo Code and Zed add ghost text alongside an agent. Cline covers the agent and chat side with any model. A team switches when completions must not leave its network, or when it wants to choose the model behind them.
3 of 4 carry an OSI licence. 4 of 4 run on your own machines. Figures read from GitHub on September 17, 2026.
One project per need, with the reason from its own README and the one thing to know before you choose it.
Its README calls it an open-source, on-premises alternative to GitHub Copilot: a server with no database or cloud service, on consumer GPUs, serving VS Code, JetBrains and Vim.
Know firstThe web server and team layer need a paid enterprise licence in production; the last release was January 2026.
MIT, fill-in-the-middle completion and chat, defaulting to a local Ollama-compatible endpoint, offline, free.
Know firstVS Code only and no agent.
MIT, inline autocomplete with tab to accept plus five agent modes, in VS Code and JetBrains, with 500+ models at the provider's rate.
Know firstNo local models described in the README.
Apache-2.0, with reviewable diffs, approval on every action, MCP, and Ollama or LM Studio for local models.
Know firstNo autocomplete described.
The facts that decide most choices. Each project's page carries the full panel.
| Project | Best for | Openness | Licence | Runs as | Own machines | Stars |
|---|---|---|---|---|---|---|
| Tabby | A self-hosted Copilot server for a team | Source available | See LICENSE | Docker (server), Built from source with Rust, VS Code, JetBrains and Vim plugins | Yes | 33,884 |
| twinny | Ghost text in VS Code from a local model | OSI | MIT | VS Code extension | Yes | 3,648 |
| Kilo Code | Completion and an agent in one extension | OSI | MIT | VS Code extension, JetBrains plugin, CLI, Web (Cloud Agent) | Yes | 27,344 |
| Cline | The chat and agent side, any model | OSI | Apache-2.0 | VS Code extension, JetBrains plugin, CLI, Desktop app, Node.js SDK | Yes | 68,575 |
Licence and star figures read from GitHub on September 17, 2026. Stars are shown as a dated fact and were not used to rank this page.
A self-hosted coding assistant server with an Apache-2.0 core and a proprietary ee directory for the web server and team features, serving code completion and chat to VS Code, JetBrains and Vim from models on your own GPU.
Its README calls it an open-source, on-premises alternative to GitHub Copilot: a server with no database or cloud service, on consumer GPUs, serving VS Code, JetBrains and Vim.
Know firstThe web server and team layer need a paid enterprise licence in production; the last release was January 2026.
An MIT VS Code extension for fill-in-the-middle code completion and sidebar chat that defaults to a local Ollama-compatible endpoint and works offline, with nine providers and no paid tier.
MIT, fill-in-the-middle completion and chat, defaulting to a local Ollama-compatible endpoint, offline, free.
Know firstVS Code only and no agent.
An MIT coding agent for VS Code, JetBrains and the CLI with inline autocomplete, Code, Plan, Ask, Debug and Review modes, an MCP marketplace, access to 500+ models with no API key to start, and a hosted cloud agent.
MIT, inline autocomplete with tab to accept plus five agent modes, in VS Code and JetBrains, with 500+ models at the provider's rate.
Know firstNo local models described in the README.
An Apache-2.0 coding agent for VS Code, JetBrains, the terminal and a desktop app that edits files as reviewable diffs, runs commands with your approval, works with any model provider including Ollama, and connects tools through MCP.
Apache-2.0, with reviewable diffs, approval on every action, MCP, and Ollama or LM Studio for local models.
Know firstNo autocomplete described.
The questions that settle it, in the order they usually come up.
No: Tabby on your own server, or twinny against a local model. Yes: any of the five.
Tabby serves both, Kilo Code and Cline have JetBrains plugins, twinny is VS Code only, Zed is its own editor.
Kilo Code has both. Cline is the agent. Zed has both in a new editor. twinny and Tabby are completion and chat.
GitHub Copilot, in the vendor's own words, is "your AI accelerator for every workflow, from the editor to the enterprise", with inline suggestions and chat in the IDE plus explanations and answers on GitHub, by its own description. github.com. The cases below are where that still wins.
Stars, licence, last commit and last release are read from the GitHub API by a script every week and carry the date they were read. Nobody types them.
What each project does is written from its own README and LICENSE, read in full on the date shown, and phrased as the project's claim. The sources are listed at the end.
That a project is faster, better or cheaper than GitHub Copilot. A price for anyone. A user count. A roadmap. If a fact is not on this page, it was not in the source.
For inline completion from a local model in VS Code, twinny (MIT). For a team completion server on your own GPU serving VS Code, JetBrains and Vim, Tabby (Apache-2.0 core, paid enterprise directory). For completion plus an agent, Kilo Code (MIT) or Zed (GPL-3.0). For the agent and chat side with any model, Cline (Apache-2.0).
Tabby, whose README describes it as an open-source, on-premises alternative to GitHub Copilot. It runs as a server with no database or cloud service, on consumer-grade GPUs or Apple Metal, and its plugins deliver completion and chat. The web server and team features sit in its enterprise directory under a paid licence.
twinny (fill-in-the-middle in VS Code), Tabby (inline completion with multiple choices through its plugins), Kilo Code (inline autocomplete with tab to accept) and Zed (edit prediction with the Zeta model). Cline is an agent and does not describe autocomplete.
twinny defaults to a local Ollama-compatible endpoint and lists offline operation as a feature. Tabby runs entirely on your server. Zed runs local models through llama.cpp, LM Studio or Ollama, though its Zeta prediction model needs sign-in. Cline uses Ollama or LM Studio. Kilo Code's README does not describe local models.
twinny and Kilo Code are MIT; Cline is Apache-2.0 with a closed JetBrains plugin; Zed is GPL-3.0-or-later with Apache-2.0 components. Tabby is split: Apache-2.0 outside its ee directory and a paid enterprise licence inside it, so this site labels Tabby source available.
Running one of these inside your own environment, with your own data and your own security rules, is the kind of work Reveneau does. Read how a forward deployed engagement works.