Claude Platform Docs

To enable the Compliance API, see the setup guide.

Set up the Compliance API

擷取遠端工作階段訊息

GET/v1/compliance/apps/sessions/remote/{claude_remote_session_id}/messages

擷取單一遠端工作階段的逐字稿：使用者提示、助理回應，以及工具呼叫和結果。不包含思考區塊和圖片。

訊息預設由最舊開始傳回；傳遞 order=desc 以反轉順序。分頁使用與清單端點相同的 page/next_page 機制，每頁最多 limit 則訊息（預設 100，最大 1000）；持續分頁直到 next_page 為 null。tool_use_input_max_bytes 和 tool_result_max_bytes 限制每個工具使用輸入和每個工具結果文字項目傳回的位元組數；因任一上限而被縮短的區塊會帶有 truncated: true。

回應會在分頁的 data 陣列旁，於 session 下嵌入工作階段的中繼資料。在此端點上，session.user.email_address 和 session.started_by_user 一律為 null；請改從清單端點讀取這些值。

若工作階段仍為 pending、已刪除，或位於金鑰可讀取之組織範圍外，則傳回 404。格式錯誤的工作階段識別碼會傳回 400。

Path parameters
claude_remote_session_id: string

The remote session identifier (cse_...) to retrieve

Query parameters
limit: optional number

Maximum results (default: 100, max: 1000)

default100
maximum1000
minimum1
order: optional "asc" or "desc"

Sort direction. asc (oldest-first) or desc.

defaultasc
One of the following:
"asc"
"desc"
page: optional string

Opaque pagination token from a previous response's next_page field. Pass this to retrieve the next page of results. Clients should treat this value as an opaque string and not attempt to parse or interpret its contents, as the format may change without notice.

tool_result_max_bytes: optional number

Truncate each text item inside a tool result to at most this many bytes (cut on a code-point boundary). Pass -1 to request the server maximum. 0 is not a valid value.

default10000
maximum2147483647
minimum-1
tool_use_input_max_bytes: optional number

Truncate each tool-use input to at most this many bytes (cut on a code-point boundary so the result is valid UTF-8). Pass -1 to request the server maximum. 0 is not a valid value.

default10000
maximum2147483647
minimum-1
Headers
"x-api-key": optional string
Returns
data: array of object{ id, content, content_unavailable, 3 more }

Transcript turns for this page, ordered by transcript position. created_at is a commit timestamp and may tie or invert under concurrent writes; do not re-sort by it.

id: string

Unique identifier for the message, e.g. csev_abc123

content: array of object{ text, truncated, type } or object{ id, input, name, 2 more } or object{ content, is_error, name, 3 more }

Content blocks within the message

One of the following:
Text object{ text, truncated, type }

Text content block.

text: string

Text content from the user or the assistant

truncated: boolean

True when text exceeded the server-defined maximum (approximately 1 MiB) and was shortened.

defaultfalse
type: "text"
defaulttext
ToolUse object{ id, input, name, 2 more }

Tool invocation requested by the assistant.

id: string or null

Tool-use ID, e.g. 'toolu_01AbC...'

input: string

Arguments passed to the tool, as a JSON-encoded string. May be shortened — see the truncated field

name: string

Name of the tool invoked

truncated: boolean

True when input was shortened. Pass tool_use_input_max_bytes=-1 to request full content, subject to the server-side maximum.

defaultfalse
type: "tool_use"
defaulttool_use
ToolResult object{ content, is_error, name, 3 more }

Result returned by a tool invocation.

content: array of object{ text, type }

Text content returned by the tool. Non-text item types are omitted.

text: string

Text returned by the tool

type: "text"
defaulttext
is_error: boolean

True when the tool reported an error

name: string

Name of the tool that produced this result

tool_use_id: string or null

ID of the tool_use block this result responds to

truncated: boolean

True when one or more text items in content were shortened. Pass tool_result_max_bytes=-1 to request full content, subject to the server-side maximum.

defaultfalse
type: "tool_result"
defaulttool_result
content_unavailable: boolean

True when the stored content could not be returned — it could not be decrypted, or it exceeded the server's per-event size bound. content is empty in that case; this distinguishes 'no content' from 'content withheld'.

defaultfalse
created_at: string

When the message was recorded (RFC 3339, UTC)

formatdate-time
role: "assistant" or "user"

Message sender (user or assistant)

One of the following:
"assistant"
"user"
sent_by_user_id: string or null

Identifier of the human account that sent this turn on an agent-owned session. Null on user-owned sessions, where every user-role turn was sent by the session's user.

next_page: string or null

Opaque page token; pass as page to retrieve the next page. Null when no rows exist after this page. Treat this value as opaque; do not parse or store it long-term, as the format may change without notice.

session: object{ id, agent_id, claude_project_id, 7 more }

Session metadata. started_by_user, user.email_address, and claude_project_id are always null on this endpoint; the messages endpoint resolves neither email addresses nor project bindings.

id: string

Remote session identifier

agent_id: string or null

Identifier of the automated agent that owns the session. Null for user-owned sessions. At most one of user and agent_id is set.

claude_project_id: string or null

ID of the project the session is bound to. Null when the session has no project binding.

created_at: string

When the session was created (RFC 3339, UTC)

formatdate-time
organization_uuid: string

UUID of the organization the session belongs to

product_surface: string or null

The Claude product the session was created from. Currently cowork_remote, for Cowork sessions started on claude.ai web or mobile. More values will appear as other surfaces launch, so treat any unrecognized value as an unclassified surface rather than an error. Null for sessions created before this field was recorded, for surfaces that do not stamp it, and for unrecognized tag values.

started_by_user: object{ id, email_address } or null

A user associated with a remote session.

id: string

User identifier

email_address: string or null

User's email address. Null when the user is no longer a member of an organization the key may read — id remains set so attribution is preserved. The messages endpoint does not resolve email addresses; this field is always null there.

status: string

Session lifecycle state. One of active, paused, archived, or failed — the lifecycle states the owning product surface exposes — plus pending, a brief transient state that resolves before any transcript content exists. The list endpoint includes pending; the messages endpoint returns 404 for it. Deleted sessions are not returned on either endpoint. Treat unrecognized values as an unknown state rather than an error.

updated_at: string

When the session was last modified (RFC 3339, UTC)

formatdate-time
user: object{ id, email_address } or null

A user associated with a remote session.

id: string

User identifier

email_address: string or null

User's email address. Null when the user is no longer a member of an organization the key may read — id remains set so attribution is preserved. The messages endpoint does not resolve email addresses; this field is always null there.

擷取遠端工作階段訊息
cURL
curl https://api.anthropic.com/v1/compliance/apps/sessions/remote/$CLAUDE_REMOTE_SESSION_ID/messages \
    -H "Authorization: Bearer $ANTHROPIC_COMPLIANCE_API_KEY"
Returns Examples
Response 200
{
  "data": [
    {
      "id": "id",
      "content": [
        {
          "text": "text",
          "truncated": true,
          "type": "text"
        }
      ],
      "content_unavailable": true,
      "created_at": "2019-12-27T18:11:19.117Z",
      "role": "assistant",
      "sent_by_user_id": "sent_by_user_id"
    }
  ],
  "next_page": "next_page",
  "session": {
    "id": "id",
    "agent_id": "agent_id",
    "claude_project_id": "claude_project_id",
    "created_at": "2019-12-27T18:11:19.117Z",
    "organization_uuid": "organization_uuid",
    "product_surface": "product_surface",
    "started_by_user": {
      "id": "id",
      "email_address": "email_address"
    },
    "status": "status",
    "updated_at": "2019-12-27T18:11:19.117Z",
    "user": {
      "id": "id",
      "email_address": "email_address"
    }
  }
}