agentic workflows · distributed S3 · no egress

Your agents read the same bytes
a thousand times a day.

On metered storage that is a bill that grows every time your agents get better at their job. Here reads are free and uncapped, so nobody has to ration retrieval to protect a budget. The endpoint runs on your own infrastructure, which means no provider is building a record of what your agents looked at, how often, or when.

Get started → See the economics
$0.006/GB-mo$0 egress11 nines durabilityS3-native
● live  ·  distributed storage network

$0.006/GB/mo
Flat, predictable storage

$0 egress
Reads don’t compound your bill

S3-native
Validated on ceph/s3-tests

11 nines
Durability, actively maintained · see the math →

The agent tax

Agentic workloads are read-amplifying. Egress is the bill.

A RAG corpus is written once and read on every query. Training data is re-read every epoch. On hyperscalers, every read is metered. On XNS, reads are free — you size for storage, not for how hard your agents work.

10 TB corpus · 30 full reads / monthXNSTypical hyperscaler
Storage (10,000 GB)$60.00/mo~$230/mo
Egress (300 TB read)$0~$27,000/mo
Monthly total~$60~$27,230

Illustrative. Hyperscaler figures from public pricing — verify before publish. XNS bills per TB of your data — the erasure-coding parity that protects it is included in the rate, not charged on top.

drop-in — change one endpoint
# your agent's boto3, unchanged
s3 = boto3.client("s3", endpoint_url="https://relayer.example.com")
s3.upload_file("./embeddings.parquet", "agent-corpus", "v2/embeddings.parquet")

Want the rest of the pipeline, not just the endpoint? Four runnable AI data lake recipes →

Your metadata

They encrypt your files. They keep your metadata. We keep neither.

Who downloads what, from where, how often — that record is worth more than the data, and 61% of government data demands target it. Every other cloud holds yours. XNS has no access to your data or your metadata. Nobody does.

Virtual Private Datacenter

Compose the datacenter your agents inherit.

A VPD is to storage what a VPC is to compute. Choose geography, durability, jurisdiction, and independence — then point any S3 tool at the result. The network is raw material; the VPD is what you ship on.

Start from the live map and build a provider set by the dimensions that matter to your workload. Watch durability, availability, and geographic spread recalculate as you go. No other cloud hands you the architect's table.

Durability Geo-diversity Jurisdiction Independence
VPD composer● composing
Durabilityhigh
Availabilityhigh
Geographic spreadmulti-region
Cost weightingoptimized
Your provider setcomposed

Machine-native

Built for agents that provision themselves.

The data plane is live today — and you can stand it up by asking your AI: the installer MCP is published now. Self-serve keys and machine-readable pricing are what we’re shipping next. Here’s what’s real and what’s on deck.

Standard S3

Every S3 SDK works unchanged. Point boto3, aws-cli, rclone, or restic at your own XNS endpoint — migrate in, migrate out.

live

Keyless OIDC sessions

Scoped, short-lived credentials via OIDC web identity — no static key to leak.

live

IAM, S3-native

Users, roles, and policies through the standard AWS IAM query API — 16 verbs. Terraform and aws-cli iam work unchanged.

live

Set up by MCP

Ask Claude to stand up your storage — npx @xns-cloud/relayer-mcp installs, registers, and verifies a Relayer for you. Set up with AI →

live

Object Lock & WORM

Write-once-read-many retention and legal holds — lock objects against deletion for a fixed window. Compliance-ready.

live

Presigned URLs

Hand out time-boxed, signed links to any object. Share or ingest without distributing credentials.

live

Built for RAG & training

Park datasets, embeddings, and checkpoints cheaply. Read them back as often as your agents demand — free.

live

Machine-readable pricing

Price and capabilities at a stable, parseable endpoint — so an agent can compare us on the merits.

coming soon

Self-serve keys

Programmatic signup and API-key issuance. No human in the loop — bucket to first byte in minutes.

coming soon

How we compare

The no-egress clouds, side by side.

For agentic workloads, architecture and egress matter more than headline storage price.

XNSAWS S3WasabiBackblaze B2
Storage / GB-mo$0.006$0.023$0.0078$0.00695
EgressNoneMeteredNone*Low / capped
ArchitectureDistributedSingle operatorSingle operatorSingle operator
Compose your own (VPD)YesNoNoNo
Agent-native control planeBuildingNoNoNo

Figures from public pricing pages · verify before publish · 2026-06 · XNS bills per TB of your data — the erasure-coding parity that protects it is included in the rate, not charged on top.

Explore the platform

The full story

The platform behind your agents’ storage — from how it works to what shipped this week.

How it works →The Relayer, erasure coding, and your Virtual Private Datacenter.S3 compatibility →Head-to-head S3 feature matrix vs the field.Eleven nines →Durability, computed — not rounded up.What's new →Everything built since you last looked.Releases →Every build, newest first — what changed and when.

Point your agents at it.

A RAG corpus. A training set. An autonomous workflow. Get keys, change one endpoint, and start storing on a network you can watch.

Get started

Read the docs

Claims on this page last verified

Product

Why XNS

Company

Legal

© Copyright - SCP, Corp | Xa Net Services and Affiliates