Amp is a coding agent built for leading models. It is available on the web, in a terminal, and on a phone, and it is designed to let agents run locally or in orbs, which are remote machines that keep working without supervision even after a laptop is closed. The service is offered to individuals on a pay-as-you-go basis with no markup.
It supports starting agents on the web, in a terminal, or from a phone, and threads can run in orbs so they can be picked up from any device. Amp is described as tuned to stay responsive on very large threads, and it includes plugins that can hook into events, add tools, standardize policy, and more. The page also says plugins can create agents, run them once, and keep talking to their threads.
Installation is available for Mac, Linux, WSL, and Windows, with Homebrew and a shell script provided. It can also be used on the web with nothing to install. The site says users can pick CPU and memory for the orbs used to run agents remotely. A news item announces four agent modes — low, medium, high, and ultra — and notes that smart, deep, rush, and large are deprecated.
In the Coding AI & assistants space, Amp takes a focused approach. It focuses on automating and managing coding workflows using AI agents directly from the terminal. Amp is a B2B product aimed at developers. Amp is paid. Amp is available on the command line.
Amp first shipped in 2026. Key capabilities include agent plugins, CLI interface, and remote control.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Closest matches by what these projects do