what is OneBucket Virtual S3?
OneBucket Virtual S3 is the AI data layer: one global endpoint and one place to control all your data, with speed and flexibility, from the first developer to the largest enterprise. It holds no data itself: it sits in front of object storage across clouds, on-premises and the edge, presents it as one logical namespace over S3 and MCP, and decides by policy where data lives, replicates and is served from. The storage underneath is unchanged.
See how it works →is OneBucket Virtual S3 a storage bucket?
No. OneBucket Virtual S3 is a virtual S3 over the buckets you already have, the way a virtual machine is not a computer: one S3 endpoint and one policy in front of storage you keep. Your objects stay in your own AWS S3, Azure Blob, GCS, Wasabi, Akave, on-premises or edge buckets; OneBucket holds no data itself. When a policy caches or replicates an object, the copy lives in a peer bucket you configured. What you get is one endpoint, one set of policies, and one place to see every call.
what does it cost, and what is free?
Sign up for s3.onebucket.io by December 31, 2026 and OneBucket™ is free for 90 days, with up to 5 TB of Wasabi storage included for the same period. You can cancel anytime. After the free period, OneBucket™ is $10 per TB per month, and storage is billed by your storage provider. The compute SDK is open source; hosted compute on Datum, our development compute partner, is priced separately. Enterprise: $120K per PB per year, the same $10 per TB per month.
See pricing → ·
Get your free account →what is the OneBucket compute SDK, and what does Datum do?
An open-source SDK that transforms data in the data path, launched with Datum, our development compute partner. All of it is on GitHub. Early development partners get early access with Datum, first come, first served. An open-source SDK to support in-line data transformation including use cases such as variable chunking, PII redaction, compression, private-key encryption and video transcoding, running on Datum compute. Use ours or write your own. It isn't part of the free offer: hosted compute on Datum is priced separately.
See the developers page →does OneBucket store or see my data?
OneBucket holds no data itself: objects live in your buckets, and a cache is a peer bucket you configure. Your backend credentials stay with their Endpoint and are never exposed to clients; applications and agents authenticate with a Core Key, inside the limits its policy sets. S3 uses AWS SigV4; MCP uses OAuth 2.1 with PKCE, and your secret key never leaves OneBucket. Every operation is written to the audit log, which you can list, fetch and tail over the events MCP server. Starting with a bucket of public data, or a starter backend from Attimis, is a fine way to try it before any company data is involved.
Privacy policy ·
Termswhich s3 clients and sdks work with OneBucket?
Any client that speaks the standard S3 API with AWS SigV4 auth: the AWS CLI, boto3 and the AWS SDKs in any language, s3cmd, rclone, and most S3-compatible tools. Point it at
https://s3.onebucket.io with your Core Key and use region
us-east-1 for signing. Both virtual-hosted and path-style addressing are supported.
See S3 API compatibility in the docs →how is this different from migrating my data to the cloud or a data lake?
Our approach is orchestrate, don't migrate. Instead of copying your data into a new system before AI can use it, OneBucket connects to data where it already lives and presents it as one logical bucket. There's no lift-and-shift project, no duplicated storage, and no rearchitecting — you keep your existing systems and policies in place.
do i need to copy or move my data?
No. OneBucket reaches your data in place. Because the Adaptive Data Fabric™ works at the API layer, it's stateless and lightweight — applying policy-driven routing and caching without forcing copies. Data moves only when your policy says so: when movement is useful (for performance or locality), it happens under the policies you define, not as a prerequisite to getting started.
which clouds and storage providers does OneBucket support?
AWS S3, Azure Blob Storage, Google Cloud, Wasabi, and Akave, plus on-premises storage, the edge, and air-gapped environments. Your applications always connect over S3 or MCP; OneBucket speaks each backend's own protocol behind the endpoint. It also connects to platforms like Snowflake and Databricks. Any provider you connect becomes part of the same unified namespace.
See Wasabi's OneBucket setup guide →does it work across regions and providers?
Yes. One OneBucket instance, in a public region or in your own data center, presents buckets from any region and any provider as one namespace. Where each bucket lives, and where it is replicated, follows that bucket's policy.
can my data stay in my own data center?
Yes. Run OneBucket on-premises or in your private cloud and map your buckets only to on-premises backends; nothing leaves. If you later want a copy elsewhere, it is one policy change, not a new pipeline.
how do ai tools and agents access the data?
Through two native interfaces:
MCP (Model Context Protocol) for AI agents and assistants, and the
S3 API for applications and tooling that already speak S3. Your AI sees one consistent endpoint regardless of where the underlying data physically lives.
See the MCP and S3 reference in the docs →what apis does OneBucket expose besides s3?
Three:
MCP for data, a second
MCP server for
events (list, fetch and tail every operation), and a
management API for keys, backends and policies. The web console sits on the same API. Once you have signed up, the full API reference is yours on request: just ask
support@attimis.co.
can we manage it as code?
Yes. Everything you set up in the console, from keys to backends to policies, can be done through the management API, so your setup can live in your own scripts and pipelines. The full API reference is available on request after sign-up.
do i have to add another platform to my stack?
No new platform sits in your tools' data path: to Databricks, Spark, DuckDB or any application, OneBucket is simply an S3 endpoint. As SaaS there is nothing to install; it is live the moment you sign up.
does OneBucket work with databricks?
Yes, for reading: Databricks reads data on-premises or across clouds through OneBucket without copying it into the cloud first. Writes managed by Unity Catalog still go to the workspace's own cloud storage, because Unity Catalog accepts only a few specific storage providers. Results can be written on-premises through OneBucket instead, and exposed read-only in Unity Catalog.
See OneBucket with Databricks →which endpoints do i connect to?
OneBucket runs in multiple regions, each exposing an
S3 endpoint and two
MCP endpoints — one for data, one for events. The
global endpoints work from anywhere — requests are routed to the live cluster nearest the caller; use a cluster-specific endpoint when you need requests pinned to a location:
- geo · global — S3 (object):
s3.onebucket.io · MCP (data): data.onebucket.io · MCP (events): events.onebucket.io - anycast · datum — S3 (object):
s3.datum.onebucket.io · MCP (data): data.datum.onebucket.io · MCP (events): events.datum.onebucket.io - us-ashburn-1 · ashburn (also us-east-1, cluster8) — S3 (object):
s3.us-ashburn-1.onebucket.io · MCP (data): data.us-ashburn-1.onebucket.io · MCP (events): events.us-ashburn-1.onebucket.io - us-sanjose-1 · san jose (also us-west-1, cluster7) — S3 (object):
s3.us-sanjose-1.onebucket.io · MCP (data): data.us-sanjose-1.onebucket.io · MCP (events): events.us-sanjose-1.onebucket.io - ap-singapore-1 · singapore (also ap-southeast-1, cluster6) — S3 (object):
s3.ap-singapore-1.onebucket.io · MCP (data): data.ap-singapore-1.onebucket.io · MCP (events): events.ap-singapore-1.onebucket.io
Client configuration for each is covered in the
docs →how does setup work?
Three building blocks: Core Keys — the provider-abstracted credential your apps use to authenticate to OneBucket; Endpoints — connections to your storage providers, each holding its own provider credentials; and Policies — a short wizard that maps Core Keys to your real buckets, with controls for read/write/delete, replication, compression, region, and sync quorum. Most teams connect a first environment in minutes.
how do people sign in?
The console supports Google sign-in today; Microsoft and GitHub sign-in are coming. Applications and agents never use these logins: they authenticate with their own keys, inside the limits their policy sets.
does my data stay secure and in place?
Yes. Provider credentials live with their Endpoint and aren't exposed to applications, which authenticate only with a Core Key. Access, replication, and movement are governed by the Policies you set — so you unlock data for AI without compromising compliance or giving up control over where data resides.
can OneBucket run on-premises or in an air-gapped environment?
Yes. OneBucket runs as SaaS, inside your VPC, on-premises, or fully air-gapped — so it fits environments with strict data-residency or isolation requirements.
what results can i expect?
It depends on your environment, but as one example, Rivet Industries used OneBucket to increase GPU utilization by 30%, drive AWS egress costs to roughly zero, and ship faster — with zero compliance compromise. The premise: less than 10% of enterprise data is reachable by AI today, and OneBucket is built to unlock the rest.
how do i get help or support?
how do i get started?