AIChat

AIChat is a command-line client for large language models. It talks to more than 20 providers through one interface, including OpenAI, Claude, Gemini, Ollama, Groq, Azure-OpenAI, VertexAI, Bedrock, Mistral, Deepseek, OpenRouter, and any OpenAI-compatible API. It runs in two modes: CMD mode for one-shot prompts, and a Chat-REPL with tab autocompletion, multi-line input, history search, configurable keybindings, and custom prompts.

Input can come from several sources. You can pipe stdin (cat data.txt | aichat), pass local files, directories, or remote URLs with -f, or pass the output of an external command (aichat -f '`git diff`'). In the REPL, the equivalent is .file. The shell assistant turns natural-language task descriptions into shell commands and adjusts them to your OS and shell.

Other features include roles (a prompt plus model configuration), sessions that keep conversation context, and macros that chain REPL commands. RAG lets you add external documents to a conversation. Function calling supports AI tools and MCP, and agents combine instructions, tools, and RAG documents. Tools and agents are built with the separate llm-functions repository.

aichat --serve starts a built-in HTTP server. It exposes OpenAI-style chat completions, embeddings, and rerank endpoints, so it can proxy multiple providers. It also serves a browser-based LLM Playground and an LLM Arena for comparing models side by side. It supports custom dark and light themes. It is aimed at developers and terminal users who want to work with several LLM providers from the command line and scripts.

Installation

# via Homebrew
brew install aichat

# via Cargo
cargo install aichat

# via Pacman
pacman -S aichat