list_items

list_items is a metadata-only tool. It reads the agent items endpoint (GET /v1/items) under the hood and returns one summary per item the calling agent has at least one permission row on. Agents with no grants get an empty array.

Input

None.

Output

A JSON object with an items array:

{
  "items": [
    {
      "id": "item_...",
      "label": "OpenAI API key",
      "storageMode": "server_managed",
      "cryptoVersion": 1,
      "contentVersion": 1,
      "profileId": "profile_...",
      "createdAt": "2026-04-12T14:00:00.000Z",
      "updatedAt": "2026-04-12T14:00:00.000Z"
    }
  ]
}

profileId is null for legacy items not yet assigned to a profile.

What’s not returned

ciphertext, serverCiphertext, encryptedItemKey, or any other encryption material
kind, tags, or notes — those live inside the encrypted payload and are not exposed by list
Field values (use use_secret or mount_secret instead)
Items the agent has no permission row on

Audit

list_items does not emit an audit row. The metadata listing is informational; the audited boundary is the access.* family (use_secret → access.mount_env; mount_secret → access.mount_file).

Why metadata only

Letting the model see secret values would defeat the entire MCP design. The model should plan with metadata (item IDs, labels, storage modes) and call use_secret or mount_secret to use the secret without seeing it.

Getting started

Tools

Security

Input

Output

What’s not returned

Audit

Why metadata only

​Input

​Output

​What’s not returned

​Audit

​Why metadata only

​Related

Input

Output

What’s not returned

Audit

Why metadata only

Related