Fine-tune, manage and deploy models without leaving your editor.
Swift · Flutter · React Native · Rust · Website
Onde Code is a coding agent for your model work on Onde Inference. It writes and runs the fine-tuning, model management and deploy code in your repo, and it works in any Agent Client Protocol (ACP) editor or in the terminal. It is free and Apache-2.0. Onde Inference is the hosted side: get an app key at ondeinference.com.
It is one static binary written in Rust, and it also accepts any OpenAI API compatible endpoint.
npm install -g @ondeinference/onde-code # pulls in the binary for your platform
npx @ondeinference/onde-code --acp # or run it without installing, from an editorPrebuilt archives for macOS, Linux and Windows are on the releases page. Building from source and the ACP registry are covered in docs/install.md.
onde-code --login # sign in with your browser and pick an Onde Inference app
cd my-project
onde-code # terminal UI--login opens ondeinference.com, and the app key comes back to the agent without appearing in a
URL. Prefer to paste a key? Run onde-code --setup.
- In an editor: Zed, Android Studio and other JetBrains IDEs, Xcode, marimo, Neovim and other ACP clients run
onde-code --acp. See docs/editors.md. - In the terminal:
onde-codeopens a chat UI that asks before every edit and command. See docs/terminal.md. - Embedded in your app: the
onde-code-agentcrate serves the agent in process, for hosts that can't launch a subprocess. - For model operations: pass
onde --mcpto the agent as an MCP server, and it can list, register and assign models on your Onde Inference account.
The docs index covers install, editors, what the agent does, the terminal UI, model workflows, providers and configuration, and development.
Issues and pull requests are welcome at
ondeinference/onde-code. Run
cargo fmt --check && cargo clippy --all-targets --locked -- -D warnings && cargo test before you
open one. docs/development.md has the layout.
Apache License 2.0. See LICENSE.
Copyright © 2026 Onde Inference