Amp

What is Amp?

Amp is a coding agent and development environment that lets developers build software from anywhere, running agents in isolated "orbs" or on their own machine via a CLI and runner. Its main differentiator is that agents keep working after the laptop is closed, so tasks can continue from a phone, overnight, or while the developer does something else.

Key Features

• Orbs: Each agent runs in its own orb, an isolated environment that keeps working after the laptop is closed. This removes the need to juggle worktrees or keep a session open to make progress.
• Amp CLI and runner: Developers can run the agent on their own machine using the CLI and runner, in addition to running agents in orbs.
• Automations and self-wakeups: Agents can schedule themselves to wake up and check on work, such as monitoring a performance rollout and comparing production numbers without being prompted.
• Production access via OIDC: Orbs can be granted or revoked access to production systems from the terminal, using OIDC and Google Cloud grants, so an agent can read production logs when needed.
• Canaries: Real agents or threads in production can be used as canaries to debug production issues and confirm whether a change produces a fleet-wide win.
• Amp-hosted projects and amp clone: New projects can be started from scratch and hosted by Amp, then cloned locally by running amp clone.
• Screen recording input: A screen recording of a bug, such as a flicker, can be sent to the agent as the basis for a fix.
• End-to-end task handling: Agents can ship changes, watch a rollout, ask for screenshot proof, cut a release, email a user, and close a bug within a single thread.

Use Cases

• Performance rollouts: An agent ships performance fixes, wakes itself on a schedule to compare production numbers, and reports whether a fleet-wide win is visible.
• Production debugging: With orb access to production logs, an agent can investigate issues and use canary threads to reproduce and diagnose problems.
• Bug reports from users: A reported bug, such as amp clone failing under Nix, can be handed to an agent that works the fix in its own thread.
• Unsupervised background work: Developers send a prompt, close the laptop, and continue from a phone or let the agent work while they sleep.
• Team collaboration: Work can be shared with teammates, and follow-up tasks found along the way can be spun up as new orbs.

Who It Is For

Amp is aimed at developers and engineering teams who want to delegate substantial coding work to agents without babysitting every session. It suits people who need agents to keep running after they step away, who want production-aware debugging with controlled access, and who prefer to avoid managing worktrees by hand. Teams that share work and monitor rollouts across a codebase get the most from the orb model.

Pricing

The source material does not state Amp's pricing model, free tier, or trial terms, so no pricing details can be confirmed here.

Frequently Asked Questions

What is Amp used for?

Amp is used to build software with a coding agent that runs in isolated orbs or on the developer's own machine. It handles tasks such as shipping performance fixes, monitoring rollouts, debugging production issues, and resolving user-reported bugs.

What is an orb in Amp?

An orb is an isolated environment that gives each agent its own workspace, so agents keep working after the laptop is closed. Orbs can be granted or revoked production access from the terminal, and follow-up tasks can be spun up as new orbs.

Can Amp agents keep working after I close my laptop?

Yes. Amp's orbs let agents continue running after the laptop is closed, and work can be continued from a phone or left running while the developer sleeps. Agents can also use automations to wake themselves and check on tasks such as production rollouts.

How does Amp access production systems?

Amp orbs can be granted access to production using OIDC, with grants managed from the terminal and revocable per orb. In the source demo, an agent used this access to read production logs and confirm a performance win.

Who should use Amp?

Amp suits developers and teams who want to delegate coding work to agents that run unsupervised and can act on production context. It is a fit for those who want to avoid managing worktrees and who value sharing agent work with teammates.