Back to services

Amp

Amp is a coding agent and development environment built for frontier models. It can run unattended on remote orb machines in the cloud, or locally on your own machine through the CLI.

View alternatives to Amp
Contributor:
einverne
Suggest edit
Recent reviews

Alternatives

Tech stack

No tech stack added yet. Sign in to suggest some

Amp is a coding agent plus a development environment, which the official site positions as "built for frontier models." Its core idea is to give every agent thread its own remote machine, called an orb. That way you can send a prompt, close your laptop, and the agent keeps working.

Key features

  • Remote agents on orbs: Every thread gets a fresh orb with your code, plugins, and tools. You no longer have to manage worktrees, you won't run into port conflicts, and there's nothing to clean up when a task ends. You can start as many orbs as you need, whether for a small fix or an entire feature.
  • Pick up from any device: View and control agents from the web, the terminal, or your phone. The official site mentions an Android version.
  • Live previews with Portals: Run your app in an orb and watch the agent's changes live through a portal, with hot reload. You don't need a local checkout or a local dev server.
  • Review and terminal: Review changes and browse the files in an orb from any device. You can also open a terminal shared with the agent, which is the same tmux session.
  • Automation and event triggers: An orb can be woken up by external events, such as a new Linear issue, a CI failure on the main branch, or a scheduled task the agent sets for itself. The official site also mentions that you can assign an issue to Amp in Linear.
  • Team collaboration: Invite teammates into the same orb to jointly control the same agent, code, and running app.
  • CLI and local use: Use the amp command in your local terminal to work with local files and tools. Use amp sync to sync an orb's changes to your local checkout. In environments with a shell, such as servers and CI, you can run the agent with amp --no-tui and then view and control it remotely from the web or your phone.
  • Extensibility: Supports MCP servers, skills, AGENTS.md, and plugins.
  • Models and subscriptions: You can link your own ChatGPT subscription so that the low, medium, and high settings of the "Dial" use that subscription and don't consume Amp credits.
  • Self-hosting and macOS: News on the official site mentions that you can self-host orbs, and that you can use macOS Satellites to build and test iOS and Mac apps.

Target users and use cases

Amp suits individual developers and teams who want to run coding agents in the cloud and work on several tasks in parallel. Typical uses include: sending a prompt and walking away while the agent works autonomously for a long time; having the agent check release status on a schedule after launch; having the agent attach screenshots or videos as proof of its changes; and reviewing changes and shipping from your phone.

Installation and deployment

  • On Mac, Linux, and WSL, the CLI can be installed with curl -fsSL https://ampcode.com/install.sh | bash. The official site also provides installation methods for Windows and Homebrew.
  • Using orbs requires an Amp account. The official site has tiered pricing; for specific prices, please refer to the official pricing page.
  • Amp is a commercial hosted service, and the official site does not state that it is open source. If you need to run it on your own infrastructure, you can use the self-hosted orb feature the official site offers.