Agents need space to run.
Fil One doesn't run up your bill.
S3-compatible object storage built for AI first.
No egress. No per-request fees.
No credit card required · 30 days free · Connects in minutes
Standard storage keeps the meter running while your agent works
Agents behave differently from any workload hyperscalers were designed for. They fetch, write, retry, and fetch again, thousands of times for a single task. With traditional S3 pricing, every operation has a cost. The bill compounds because the work compounds.
Most teams end up deleting the data their agents generate because it costs too much to keep. The valuable context, history, and outputs your agent built up last month, gone.
Fil One changes the model. Let your agents experiment and keep what they generate.
| Cost item | AWS S3 | Fil One |
|---|---|---|
| Storage | $0.023/GB/month (~$23/TB) | ~$0.005/GB/month ($4.99/TB) |
| PUT requests | $0.005 per 1,000 | Included |
| GET requests | $0.0004 per 1,000 | Included |
| Egress | $0.09/GB | $0 |
| Predictable bill | No | Yes |
AWS S3 Standard pricing as of May 2026, us-east-1. 1 TB = 1,000 GB (decimal). Actual AWS costs vary by tier, region, and volume discounts.
Your existing S3 code works
Just swap the endpoint, and your agents are running on Fil One. Your boto3, AWS CLI, and SDK calls work as-is, and buckets behave like you expect. So you can start running in minutes.
import boto3
s3 = boto3.client(
"s3",
endpoint_url="https://eu-west-1.s3.fil.one",
aws_access_key_id="YOUR_ACCESS_KEY",
aws_secret_access_key="YOUR_SECRET_KEY",
)
# Upload a file (create the bucket first in the dashboard)
s3.upload_file("my-file.txt", "my-sdk-bucket", "my-file.txt")
# Verify
response = s3.list_objects_v2(Bucket="my-sdk-bucket")
for obj in response.get("Contents", []):
print(f"{obj['Key']} ({obj['Size']} bytes)")Built differently. Built for agents.
Know what you'll pay upfront
No per-call charges. No egress. You can read the pricing page once and stop thinking about it. Run your agents as hard as you need.
Go all in without getting locked in
With no proprietary format, vendor-specific API, or migration tax, you can move your data and workloads freely. No surrendering your agents to a model you can't get out of.
Storage your agents can rely on
Routed across independent infrastructure providers worldwide. Distributed and redundant by design. Backed by audit-ready visibility into storage integrity.
What agents need from storage
Start building today. Be the first to test upcoming features.
Persistent agent memory and artifacts
Your agents write session state, conversation history, checkpoints, and generated outputs straight to S3 buckets, and read them back on the next run. Flat pricing means the loop doesn't cost you: thousands of small reads and writes price the same as a handful. Keep everything your agents produce instead of deleting it to stay in budget. Plain S3: works today with boto3, the AWS CLI, or any SDK.
A storage tool your agent can call directly
Give your agent read/write/list access to your buckets without hand-writing the S3 glue. The Agent Toolkit exposes storage as a native tool you drop into Claude Desktop, Cursor, or your framework via MCP, so the agent reaches storage through a clean interface, and the data stays in buckets you own, not a third-party SaaS.
A queryable knowledge base over your buckets
Point a retrieval agent at a bucket and query it in plain language. Files index as they land; semantic search runs on your own model keys. Store the whole corpus at volume — flat pricing makes a large document set a storage cost, not a per-query tax.
Free trial: 1 TB storage + 2 TB bandwidth, 30 days.
No credit card required.
Ready to get the most out of your agent storage?
Talk to us about your use case, or head to the docs and start building today.