Review bucket configuration
Policies, ACLs, encryption, versioning, and public-access settings — flagged when a combination is risky.
Native desktop agent
A native agent for object storage
Point it at your S3-compatible storage and ask — review config, analyze access logs, or triage a failing request. Keys and evidence stay on your machine.
Which bucket, log, or error should we check?
Read-only and never destructive — I'll ask before moving any data.
Ask Storage Agent… ( / for commands )
your model
Diagnose an errorAnalyze access logsInventory & capacityReview bucket configMap account & bucketsWhere cost concentrates
Native desktop agent
Ask about your storage, get evidence back
Start a chat or pick a task — bucket config, access logs, inventory, errors, or cost. The agent plans read-only checks, works in DuckDB on imported evidence, and cites the log rows or config lines behind each finding.
What it investigates
What it investigates
Analyze access logs
Import access logs and query them in DuckDB: who read what, hot prefixes, and unusual access.
Inventory and capacity
Turn an inventory into answers about object counts, sizes, storage classes, and growth.
Map accounts and buckets
Sort out which buckets exist across accounts, what each is for, and which look orphaned.
Triage errors
Bring a failing request; it checks the object with range reads and metadata to narrow the cause.
Optimize cost and storage
See where cost concentrates and where lifecycle or tiering would actually help.
How a session runs
How a session runs
Ask in plain language
Start a chat, or pick a task like review bucket config or analyze access logs.
The agent plans
It picks read-only tools, reads policy, metadata, and logs, and works in DuckDB.
Evidence, not guesses
Findings come with the log rows, object metadata, or config lines they rest on.
You decide
Read the diagnosis, approve anything that moves data, and act with the evidence in hand.
Local and read-only
Local and read-only
Keys in a local vault
Credentials live in an encrypted vault on your machine and are never written to logs, reports, or model prompts.
Read-only cloud access
The agent can inspect but not write: no deletes, no overwrites, no configuration changes.
You approve data movement
Downloads, large scans, and dataset analysis always ask first — nothing leaves without your go-ahead.
FAQ
Which AI model does it use?
You bring your own: configure a model provider and key. The key stays in the local vault and is never sent anywhere but your chosen provider.
Does my data leave my machine?
No. Credentials, imported evidence, and analysis stay local; any action that moves data asks for consent first.
Can it change my storage?
No. Cloud access is read-only by design — no writes, deletes, or configuration changes.
How is this different from StorageOps?
StorageOps gives an existing agent (Pi) storage skills; Storage Agent is a self-contained desktop app that runs the whole investigation itself.
What platforms does it run on?
A Tauri desktop app for macOS, Windows, and Linux. The project is Apache-2.0.