`list_objects`¶

List one folder level of a cloud object store by URL: Amazon S3 (and S3-compatible providers), Azure Blob Storage, or Google Cloud Storage. Read-only analytics (stays available under --mcp-read-only).

When to use¶

"What files are in s3://reports/2026/?"
Browse a bucket before opening a file with read_table.

Input schema¶

Parameter	Type	Required?	Default	Description
`url`	string	yes	(no default)	`s3://bucket/prefix`, `az://container/prefix`, `gs://bucket/prefix`, `sftp://user@host/folder/` or `davs://host/folder/` (empty prefix = bucket root)

Credentials¶

The MCP/CLI server authenticates with ambient credentials:

S3: AWS_* environment variables or a cached SSO session.
Azure: an Azure CLI login, plus AZURE_STORAGE_ACCOUNT (an az:// URL cannot carry the storage account name).
GCS: Google application-default credentials, or GOOGLE_* environment variables.
SFTP: your SSH agent, as the URL's user, with strict known_hosts checking. A saved connection supplies a key file or password instead.
WebDAV: only through a saved connection (the URL cannot carry the password).

The in-app assistant additionally uses your saved cloud connections (Settings > Cloud storage), so it can reach buckets configured with static keys or per-connection sign-in.

Response shape¶

{
  "url": "s3://reports/2026/",
  "count": 2,
  "objects": [
    { "name": "q1", "key": "2026/q1/", "url": "s3://reports/2026/q1/", "is_folder": true, "size": null, "modified": null },
    { "name": "summary.parquet", "key": "2026/summary.parquet", "url": "s3://reports/2026/summary.parquet", "is_folder": false, "size": 81234, "modified": "2026-06-20T11:02:00+00:00" }
  ]
}

Open a listed file by passing its url as the path of any read tool.

Example call¶

{
  "name": "list_objects",
  "arguments": {
    "url": "s3://reports/2026/"
  }
}

list_objects¶