# Glac

> Glac is a compute platform for developers, teams and the AI agents they build. Any container image runs in a Firecracker microVM of its own, which sleeps when idle and wakes in 3.5 ms with its memory, processes and connections intact. CPU is billed by the busy millisecond and memory while awake; asleep, an instance pays only for what it keeps.

Every product runs on Compute: instances, each a microVM, made from any OCI image, with a key (named, keeping its memory and disk) or without (one of many, picked for each request). Data centres running now: Ashburn (North America) and Coventry (Europe); being built: New York (North America), Frankfurt (Europe), Finland (Europe), Singapore (Asia-Pacific) and Japan (Asia-Pacific). Measured medians, on bare metal (AMD Ryzen 9 9950X, 3 October 2026), nothing else running: 3.5 ms to wake, 2.3 ms to sleep; how they were measured is on https://glac.ai/benchmarks/.

## Products

- [Cube](https://glac.ai/cube/): Cube. Any container image as an app that sleeps between requests. Docs: https://glac.ai/docs/cube/
- [Volume](https://glac.ai/volume/): Cube. Cube’s disks, which outlive their instance. Docs: https://glac.ai/docs/drive/
- [Sandbox](https://glac.ai/sandbox/): Managed Applications. A computer of its own for every agent. Docs: https://glac.ai/docs/sandbox/
- [Browser](https://glac.ai/browser/): Managed Applications. Real browsers, driven from code (coming soon).
- [Serverless Postgres](https://glac.ai/database/): Managed Applications. Postgres that sleeps when idle and wakes on a connection (coming soon).
- Third-party apps: Managed Applications. Open-source projects run for you, then a marketplace (coming soon).

## Solutions

- [AI agents](https://glac.ai/solutions/ai-agents/): A computer for every agent, 24/7.
- [Vibe coding](https://glac.ai/solutions/vibe-coding/): Run every app your users prompt.
- [Web apps and APIs](https://glac.ai/solutions/web-apps/): Any app, asleep until it is called.
- [Per-user backends](https://glac.ai/solutions/per-user-backends/): An instance for each user.
- [Browser automation](https://glac.ai/solutions/browser-automation/): Real browsers, driven from code (coming soon).
- [Platforms](https://glac.ai/solutions/platforms/): Give your users computers of their own.
- [Enterprise](https://glac.ai/solutions/enterprise/): Isolation, regions, contracts.

## Pricing

- [Pricing](https://glac.ai/pricing/): the one price list, the two plans and every limit of each, a calculator for Cube and each Managed Application, how billing works. Prices in US dollars; not yet registered for GST.
- One price list: Cube and every Managed Application (Sandbox, Browser, Serverless Postgres) are billed by the Cube resources they use, at these prices.
- Plans: Free, with no card, a quota of each meter every day (back at 00:00 UTC), never charged; and Paid, $5 a month, with a quantity of each meter included every month and the list price past it.
- What is kept: What is kept is priced by the GB-month, and shown by the millisecond, the second and the hour for a 730-hour month: $0.02 a GB-month for what sleeping instances keep (their memory image and the part of their disk that is their own; templates are free), snapshots, and a volume's data, all of it, in tiered storage; $0.06 more for each GB of a volume kept on the machine's NVMe, the size you set (all of it: $0.08 a GB-month; none: $0.02, at network speed); $0.02 for logs kept, as printed (request and platform lines are free).

### Cube resources

- CPU: $0.01 a million CPU-ms. Busy CPU time, by the millisecond
- Memory: $0.00222 a million GiB-ms. Held, by the millisecond, in 128 MiB steps
- Requests: $0.30 a million. Per request into a Cube or through a sandbox link
- Traffic out (NA · EU): $0.004 a GB. Per GB leaving North America and Europe
- Traffic out (APAC): $0.01 a GB. Per GB leaving Asia-Pacific
- Kept asleep: $0.00000761 a million GB-ms ($0.02 a GB-month). What sleeping instances keep, and snapshots

### Volumes

- Data: $0.00000761 a million GB-ms ($0.02 a GB-month). All a volume holds, kept in tiered storage
- NVMe cache: $0.0000228 a million GB-ms ($0.06 a GB-month). The part on the machine’s NVMe, the size you set; no write or read fees

### Logs

- Logs: $0.00000761 a million GB-ms ($0.02 a GB-month). What Cubes print that is kept, as printed

## Docs

- [Docs for language models](https://glac.ai/docs/llms.txt): the docs' own llms.txt, page by page, with the API reference.
- [The whole docs, one file](https://glac.ai/docs/llms-full.txt)
- [Quickstart](https://glac.ai/docs/quickstart/): a key, then a sandbox from the command line or from Python.

## Optional

- [How we measure](https://glac.ai/benchmarks/): sleep, wake, snapshot and CPU, with the machines, the days, the method and the raw results.
- [Blog](https://glac.ai/blog/), with its feed: https://glac.ai/blog/rss.xml
- [Changelog](https://glac.ai/changelog/), with its feed: https://glac.ai/changelog/rss.xml
- [Help](https://glac.ai/help/): search, status, how to reach us, common questions.
- [Status](https://glac.ai/status/), [About](https://glac.ai/about/), [Security](https://glac.ai/security/), [Contact](https://glac.ai/contact/)
- [Terms of Service](https://glac.ai/terms/), [Privacy Policy](https://glac.ai/privacy/), [Acceptable Use Policy](https://glac.ai/acceptable-use/)