Skip to content

Object storage

An S3 bucket connects as a browsable connection, listed alongside your servers and opened like any other. It covers the object-storage work that sits next to operations rather than next to application code: pulling a release artifact, reading an archived log, checking that last night's backup landed.

Connecting a bucket

Choose S3 (AWS S3) as the connection type. The form drops the host, port and username fields — object storage has no shell — and asks for these instead:

Field Notes
region Defaults to us-east-1 if you leave it blank
bucket The bucket this connection opens
access key ID
secret access key Encrypted by the operating system, like a password
custom endpoint (optional) For an S3-compatible service; see below

The connection opens as a tab with two panes rather than a terminal: your own computer on the left (Local) and the bucket on the right. Drag files between them to upload or download. Where object-storage support is unavailable in the installation, the connection reports that rather than failing obscurely.

During the free trial and without a subscription, an S3 tab counts toward the 10 sessions you can have open at once; see Free trial and limits.

Connect a bucket, step by step

  1. Click + new connection and choose S3 as the type.
  2. Give the connection a name, and fill in region, bucket, access key ID and secret access key.
  3. For a service other than AWS, fill in custom endpoint (optional).
  4. Leave Save connection on and click connect.

S3-compatible endpoints

Filling in custom endpoint points the client at that URL instead of AWS and switches it to path-style addressing, which is what self-hosted gateways expect. That makes MinIO, Ceph object gateways and other S3-compatible services reachable with the same connection type — the credentials and the bucket work the same way, only the endpoint differs.

Leave the endpoint blank for AWS itself.

Browsing a bucket

A bucket is presented through the same explorer as an SFTP filesystem. Key prefixes appear as directories and objects as files, so you navigate, open, edit, upload, download, rename, copy and delete in the interface you already know. Listings are paginated, so a bucket with a large number of objects browses one level at a time rather than trying to enumerate everything.

Check both of these before acting on a production bucket:

  • A "folder" is a prefix. Creating one writes a zero-byte object whose key ends in /, which is the same convention the AWS console uses.
  • A rename is a copy followed by a delete, and renaming a prefix does that for every object beneath it. It is not an atomic operation, and it is not free on a large prefix.

There is no terminal for an S3 connection, so there is no scrollback for the assistant to analyse, and AI works only in SSH and Local tabs. To show the assistant an object, download it and attach it in an SSH or Local tab's AI panel with + → Upload from computer. An attachment goes through the same redaction layer as any other file — see AI and files.

See also