2. Quickstart

First run (interactive)

rayu

On first launch you'll go through a short setup:

1. Theme — pick a color theme.
2. Provider setup — choose a provider (Anthropic, NVIDIA, DeepSeek, Kimi/Moonshot, Doubleword, OpenAI, OpenRouter, or a local endpoint) and paste your API key. For local/custom endpoints you also enter a base URL and a default model.
3. Trust — confirm you trust the current working directory (Rayu can read, edit, and run files there).

Rayu then fetches the provider's model list and drops you into the chat REPL.

Already have your key in a .env file? Rayu auto-imports known keys on startup — see Providers.

Your first conversation

Type a prompt and press Enter:

> explain what this project does and list its main modules

Useful in-session commands (type / to see all):

Press Esc to cancel a running turn; Ctrl+C twice to exit.

Headless / scripted use (print mode)

Run a single prompt and print the result (no TUI):

rayu --print "write a one-line summary of package.json"

With explicit provider + model (no saved config needed):

RAYU_OPENAI_COMPATIBLE=1 \
RAYU_OPENAI_BASE_URL=https://integrate.api.nvidia.com/v1 \
RAYU_OPENAI_API_KEY=nvapi-xxxxx \
rayu --print --model meta/llama-3.3-70b-instruct "summarize this repo"

JSON output for scripts:

rayu --print --output-format json "list top-level modules"

Auto-approve tool use (sandboxes/CI only — see security note in CLI Reference):

rayu --print --permission-mode bypassPermissions "read README and summarize"

Pick a chat model

When choosing a model, prefer instruction/chat models (e.g. meta/llama-3.3-70b-instruct, deepseek-chat, deepseek-ai/deepseek-v4-pro). Base/code/embedding/OCR models (codegemma, *-embedding, *-ocr, starcoder) are not chat models and will return 404 on the chat endpoint. See Troubleshooting.

Next: Providers →