Dedalus Labs
Fast, persistent Linux sandboxes for AI agentsStart stateful AI agents in about 50ms
Highlights
What you can do with it
- Run and test AI agents with storage that persists
- Deploy enterprise apps with fast startup and live migration
- Run large-scale research computations and ML workloads
How to use it
- 1 Install the Dedalus CLI with the command for your operating system
- 2 Set your API key as an environment variable
- 3 Create a machine with your chosen CPU, memory, and storage
- 4 Run commands on the machine or connect over SSH
- 5 Hibernate the machine when idle to stop compute charges
Who it is for
Pros
- Very fast startup shortens iteration time
- Strong VM-level isolation for safer multi-tenant use
- 50 free compute hours per month to start
Limitations
- Free hours are limited and pricing may change
- The CLI and API workflow needs some engineering comfort
- Storage is billed separately, so total cost adds up
What Dedalus Labs is
Dedalus Labs is a company backed by Y Combinator. Its product, Dedalus Machines, provides full Linux virtual machines built as computing infrastructure for AI agents. The headline feature is speed: machines start in about 50ms because the system uses snapshot recovery instead of pulling a container image. Isolation is at the VM level, and each machine has its own independent guest kernel, with the host CPU enforcing boundaries. Storage is persistent and POSIX-compliant, and memory volumes and compute are kept after a restart. Storage and compute are separated, and the system supports real-time virtual machine migration with zero downtime. The API is small, with just create, execute, and hibernate. You get 50 free compute hours per month, and the professional plan starts at $20 per month, billed by the second, with storage priced separately.
What you can do with it
- Start a full Linux VM in about 50ms and run any Linux program
- Keep state across restarts with persistent, POSIX-compliant storage
- Migrate a running machine with no downtime
- Control everything through a three-call API or SSH
Who it is for
- AI developers who need fast, stateful agent sandboxes
- Small tech companies that want quick, low-cost deployment
- Research institutions running large-scale computations
What to watch out for
The free tier gives 50 compute hours per month, and the pricing and limits may change over time, so confirm current terms before you build on it. Storage is billed separately, so your real cost is more than the compute rate alone. The workflow is CLI and API based, so it suits people comfortable with the command line and some engineering setup. Because it is infrastructure, plan for how you will manage keys, access, and costs as usage grows. Track your usage from the start so a small test does not turn into a large bill, and hibernate machines when they are idle to avoid charges.
FAQ
How fast do machines start?
Dedalus Machines start in about 50ms using snapshot recovery instead of pulling an image.
Is storage kept after a restart?
Yes. It provides a persistent, POSIX-compliant file system, and memory volumes and compute are retained after restart.
Is there a free tier?
Yes. You get 50 hours of free compute time per month, and the professional plan starts at $20 per month.
Why it is free
What the official page says： 免费不收费
Free limits, trial length and features are decided by the official site and can change at any time.
Product preview
Screenshot archived from the listing page, for preview only.
