rauta-ai

Terminal controls for Rauta models, active GPU compute, spend, and local machine setup.

npx rauta-ai

The package installs both rauta and rauta-ai. Use rauta after installation.

Rauta opens its account and control panel without requiring a coding client. Set up your account and choose the clients you want to connect:

npx rauta-ai init --clients pi,codex,grok
# Account only; install a coding client later:
npx rauta-ai init --clients none

Claude Code, Codex, Grok Build and Pi keep independent configurations. Refresh Pi's Rauta catalogue with rauta-ai pi refresh. Client status distinguishes configuration from an installed, usable client.

For Claude Code, launch a scoped Rauta session:

npx rauta-ai claude

With a supported existing Anthropic Messages-compatible gateway, the native model picker combines its models with Rauta's. Selecting a model routes that request to its destination; credentials stay bound to that destination. Ordinary Claude settings remain unchanged. Selecting Rauta sends the included conversation history to Rauta. Native Bedrock/Vertex/Foundry modes and unverified managed, credential-helper or custom-certificate setups require a separate integration.

Existing-gateway usage stays local unless you explicitly launch with rauta-ai claude --share-usage. Reporting includes tokens, not prompts or private model names. Unknown costs remain unknown. No failed model request silently switches to another provider; choose another model explicitly.

This package is a thin launcher: it downloads the static Go binary for your platform (darwin/linux, arm64/amd64) once per version. It prefers rauta-<os>-<arch> on rauta.ai/cli, then falls back to infra-arena-* and/or rauta.ai, caches it, and passes through.

curl -fsSL https://rauta.ai/cli/install.sh | sh

Docs: https://rauta.ai/docs · Rauta Labs, Inc.