Introduction

Aether is a modular, open-source AI coding agent toolkit written in Rust.

Design principles

• Empty by default — Agents have no default system prompt or tools. You add what you need.
• Any model — Use any LLM you want, local or remote. Anthropic, OpenAI, OpenRouter, DeepSeek, Gemini, Moonshot, ZAI, Llama.cpp, Ollama, and more are supported out of the box.
• Any tools — Agents get tools via MCP servers. You can author tools in any language.
• Any surface — Run your agent in a TUI, IDE/editor (via ACP), or headless.

Getting started

1. Install

   Homebrew

   brew install jcarver989/tap/aether

   Shell installer

   curl --proto '=https' --tlsv1.2 -LsSf https://github.com/jcarver989/aether/releases/latest/download/aether-agent-cli-installer.sh | sh

   Cargo

   cargo install aether-agent-cli

   Requires Rust 1.85+ — install via rustup if you don’t have it.

   Build from source

   git clone https://github.com/jcarver989/aether.git
   cd aether
   cargo build --release

   Binaries are written to target/release/aether and target/release/wisp.

2. Create your first agent

   cd your-project
   aether agent new

   ✓ Created .aether/settings.json
   ✓ Created .aether/mcp.json
   ✓ Created .aether/DEFAULT.md
   ✓ Created AGENTS.md

   • Directory.aether/

     • settings.json — Agent definitions
     • mcp.json — MCP server config
     • DEFAULT.md — Default prompt scaffold
   • AGENTS.md — Project instructions

   Built-in in-memory servers you can enable in .aether/mcp.json include coding, lsp, skills, tasks, subagents, survey, and plan.

   Set up your LLM credentials

   Most providers need an API key (for example export ANTHROPIC_API_KEY=sk-ant-...). See LLM Providers for all supported providers and credential setup.

3. Run it

   Interactive TUI

   aether

   The default experience — a terminal interface for conversations with your agent.

   IDE / editor (via ACP)

   aether acp

   Exposes agents over ACP for editor integration.

   Headless

   aether headless "Explain the main function in this project"

   Run a single prompt and exit — ideal for scripts and CI. See Headless CLI for all flags and output formats.

Next steps

Agent Settings Define agents, models, and tool access

MCP Servers Configure which tools are available

LLM Providers All supported models and credentials

System Prompts Customize agent instructions with markdown