About Ephemeral
Ephemeral runs AI agents in disposable machines. Each task starts a fresh, sealed microVM with only the tools, files and budget you allow, streams every step, and is destroyed when the work is done.
What Ephemeral is
Ephemeral is a service and a command-line tool for running AI agent work in the background. You describe a task: a prompt and input, the files, connections and websites it may use, a model and a spending limit. Ephemeral runs it on request or on a schedule, records the timeline and the exact cost, and can deliver the result with a signed callback.
Why it exists
Agents are useful because they act, and risky for the same reason. A long-lived server for an agent means leftover files, shared credentials and costs nobody capped. Ephemeral takes the opposite approach: every run is isolated in its own machine, credentials never enter that machine, connections are limited to the methods you allow, the budget is a hard limit reserved before the run starts, and nothing outlives the run except the record you choose to keep. With zero data retention, not even that.
Who it is for
Developers and teams who add background AI work to their products and operations: investigating a failed payment for support, watching a page or a metric, writing a scheduled report, processing data in a sandbox, or running code you do not trust.
The hosted service
The hosted service at ephemer.al runs in Japan. You pay from prepaid credits, so a run can never cost more than the budget you set. Availability is public on the status page.
The self-hosted edition
If your data must stay in your own cloud account, you can run the self-hosted edition under a license. It is the same product, with the same API and command-line tool, on infrastructure you control.