Introduction

clai is a command-line interface that integrates multi-vendor AI models directly into your terminal workflow. Built as a fast, composable Go binary, clai follows unix philosophy: pipe stdin into prompts, stream responses to stdout, chain tool calls over the Model Context Protocol (MCP), and manage conversations without leaving your shell.

Core commands

Command	Shorthand	Description
clai query <prompt>	q	Run a one-shot query with arguments and optional piped stdin.
clai chat	c	List, continue (c continue <id>), or delete saved conversations, or import transcripts from external coding agents.
clai photo <prompt>	p	Generate images with configured photo models (-pm, -p-dir, -p-prefix).
clai video <prompt>	v	Generate videos with configured video models (-vm, -v-dir, -v-prefix).
clai setup	s	Interactive TUI wizard for configuring text/photo/video models, profiles, and shell contexts.
clai skills	sk	List and invoke Agent Skills (SKILL.md) from user and repo skill directories.
clai shell_context	sc	Inspect and render shell context templates that inject OS, shell, and git metadata.
clai version	v	Print the installed clai version and build information.

Supported LLM vendors

clai supports major cloud and local model providers out of the box:

• Anthropic (Claude Opus, Sonnet, Haiku)
• OpenAI (GPT and reasoning models)
• Google (Gemini models)
• Mistral, DeepSeek, XAI (Grok), and OpenRouter
• Ollama for local offline models

See Supported vendors and Configuration for vendor setup details.

The community hub

clai.chat hosts a public hub for the parts of a setup that are worth copying: profiles, MCP servers, shell contexts, SKILL.md skills, themes and prompts. Anyone reads and downloads a config; a signed-in reader rates it from one to five stars.

See Share and rate configs and Quickstart to get started.