Answers to common questions about what Shelby is, what it is not and how it works today.
What is Shelby?
Shelby is a verifiable data platform for distributed compute. Workloads reach the same data through the same S3-compatible interface across supported Shelby locations, with fewer full copies to manage, control over where data is placed and a cryptographic record created at every write.
Who is Shelby for?
AI builders, platform teams and infrastructure providers working across distributed compute. It is most useful when workloads move between environments, when the same dataset serves several workloads, or when data is collected, processed, trained on or served in different places.
Is Shelby just object storage?
Shelby uses an S3-compatible object storage interface, but the problem it solves goes further: what happens to data access, copies, placement and write records when compute changes.
How is Shelby different from AWS S3 and other cloud storage?
Cloud object storage works well when data and compute live in the same cloud, and if yours will stay that way, it may be the right choice.
Shelby is built for when they do not. It uses the same S3 interface, so existing tools still work, but it is designed for workloads whose compute moves between providers, regions or environments. Workloads reach the same data across supported Shelby locations, with fewer full copies to manage and no per-gigabyte charge for moving data between those locations. Shelby also creates a cryptographic record at every write, so an authorised party can check what was committed and when.
Is Shelby a compute marketplace or aggregator?
No. Shelby does not source or broker compute. It sits at the data layer, so that where your data lives does not limit where your workloads can run.
Is Shelby decentralized?
Shelby is distributed storage. The architecture supports distribution across independent machines and locations. Testnet showed that the useful level of distribution and coverage depends on the workload, so the network model is being built around those requirements.
What is Private Beta?
Private Beta is Shelby's current controlled customer-onboarding stage. Select customers are running real data workloads while we build deployment patterns, operating processes and the next set of product capabilities around their requirements.
How do I get access?
Talk to us.
Can I use Shelby with existing tools?
Yes. Shelby supports the S3 interface teams already use, so existing S3 tools and SDKs work against a Shelby endpoint. That does not mean every migration is effortless, and Shelby does not provide file system semantics such as POSIX or NFS. We will publish a full compatibility matrix; until then, ask us about the operations your pipeline depends on.
What does "Single Address, Global Access" mean?
Workloads reach the same data through the same interface across supported Shelby locations, so the access model stays consistent as compute changes. "Global" describes the design, not a claim that every cloud, provider or location is supported.
Where is my data stored?
Only in the Shelby locations you approve. You choose which supported locations can hold your data, and placement is set by you rather than by where your compute happens to run.
How does resilience work?
Shelby uses erasure-coded placement. Data can be recovered from surviving fragments, without keeping another complete replica just for recovery.
Does data still move?
Yes. Data still has to reach the place it is used. Shelby reduces repeated full-copy staging and synchronization; it does not eliminate physical data movement.
Is Shelby zero-copy?
No. Shelby means fewer full copies to manage, not zero copies.
Why do copies matter?
A copy costs more than storage. Each additional full copy can bring its own sync job, permissions, versions to reconcile and data path to operate. Shelby reduces the number of independent copies that have to stay aligned.
What is the cryptographic record?
A sequenced cryptographic record is created as part of every write. An authorized party can check which object or version was committed, when it was written and in what order writes occurred. The record is tamper-evident by design.
What is the pricing model?
Shelby is priced on capacity. Movement between Shelby locations is not charged per gigabyte, and egress to the public internet is quoted separately alongside the capacity rate.
What is available today, and what is on the roadmap?
Today, Shelby provides Single Address, Global Access through an S3-compatible interface, fewer copies to manage, placement control, erasure-coded resilience and a cryptographic record at write. We are exploring future directions and will share more via demos or early releases.
How is Private Beta different from Testnet?
Testnet was designed to validate Shelby's architecture in an open environment and understand how builders interacted with it. Private Beta starts from that foundation and puts Shelby into real customer data paths, so customer requirements can shape deployment patterns, operating processes and the product roadmap.