Download · v0.3.0
Get Larik
One binary, no runtime to install. Pick your platform, or build it from source in a few seconds with Go 1.27 or newer.
Build from source
You need Go 1.27 or newer. From a checkout of the Larik repository:
go build -o larik ./cmd/larik
./larikTo run it from anywhere, put the binary on your PATH:
install -m 755 larik ~/.local/bin/larikFirst run
Start larik in a project directory. With nothing configured, a setup wizard finds a running Ollama or LM Studio server and any API keys in your environment, tests the connection, lists the provider's models and saves your choice. Run /connect any time to add another provider, including NVIDIA NIM.
If you already have a key exported, Larik uses it right away:
| Environment variable | Default model |
|---|---|
ANTHROPIC_API_KEY | anthropic/claude-opus-5 |
OPENAI_API_KEY | openai/gpt-5.5 |
GEMINI_API_KEY | gemini/gemini-3.8-flash |
No key? Use a local model through Ollama or LM Studio, or /connect codex to sign in with a ChatGPT plan.
Optional extras
- Linux sandbox: install bubblewrap (
bwrap) so shell commands run confined. macOS uses the built-in Seatbelt. - Windows shell: install Bash (for example, Git Bash) for the
bashtool. Windows has no built-in Larik sandbox, so shell commands ask for approval. - Local models: Ollama or LM Studio. The wizard can download a recommended Ollama model for you.
- Language servers such as
goplsortypescript-language-server, for diagnostics after every edit.