List activities
Individual model and tool calls, newest first.
Lists activities under an organisation, most recent first by default.
Request
GET /v1/organisations/{organisation}/activitiesGET /v1/organisations/{organisation}/workspaces/{workspace}/activities
Parameters
| Name | In | Type | Required | What it is |
|---|---|---|---|---|
pageSize |
query | integer | no | Maximum activities to return. |
pageToken |
query | string | no | A page token from a previous response, to retrieve the next page. |
filter |
query | string | no | Keeps only matching calls: terms of field = value joined by AND, e.g. user = "8c21e0b4". |
orderBy |
query | string | no | Sort order. |
startTime |
query | timestamp | no | Start of the window to list, inclusive. |
endTime |
query | timestamp | no | End of the window to list, exclusive. |
Response
| Field | Type | What it is |
|---|---|---|
activities |
list of object | The activities in this page. |
nextPageToken |
string | Token to retrieve the next page, empty when there are no more. |
Each item of the list
| Field | Type | What it is |
|---|---|---|
name |
string | The resource name of the activity. |
agent |
string | The agent that performed the work. |
request |
string | Opaque identifier of the end-user request this activity belongs to, as the caller's own tracing names it. |
user |
string | Opaque identifier of the person the work was done for. |
model |
string | The model that was called, as the provider names it, e.g. gemini-2.5-pro. |
estimatedCostMicros |
int64 as string | What the call is expected to cost, in millionths of a US dollar. |
status |
enum | How the call ended. |
errorCode |
string | The provider or platform error code, where the call did not succeed. |
durationMs |
int64 as string | How long the call took, in milliseconds. |
occurredAt |
timestamp | When the work finished, as reported by the recorder. |
session |
string | Opaque identifier of the conversation or session, where the product has one. |
callerService |
string | The service that made the call, e.g. atlas-agent, deep-research. |
callerComponent |
string | The part of the calling service that spent the money, e.g. chat_turn, tool:web_search, research_merger. |
subAgent |
string | The agent inside a multi-agent run that made this call, as the framework names it, e.g. meetings_researcher: the specialist a parallel or delegating agent handed the work to, or the root agent where nothing was handed on. |
environment |
string | The environment the call was made in, e.g. production or staging. |
skill |
string | The area of the product the person was using, e.g. canvas, doc_viewer, social. |
billedBy |
string | The provider that billed for the call, as a stable uppercase token, e.g. VERTEX_AI, ANTHROPIC, OPENAI, PERPLEXITY, DEEPSEEK. |
usageFormat |
string | The reporting convention the counts on this record arrived in, e.g. VERTEX, ANTHROPIC, OPENAI_CHAT, OPENAI_RESPONSES, OTLP_GENAI. |
provider |
enum | Deprecated: use billedBy, which is not limited to the four providers this enum can name. |
promptTokens |
integer | Tokens in the prompt, excluding any served from or written to the cache. |
candidateTokens |
integer | Tokens the model generated and was billed for at its output rate, including any it spent reasoning. |
cachedTokens |
integer | Prompt tokens served from the provider's cache, billed at the cache read rate. |
cacheWriteTokens |
integer | Prompt tokens written to the provider's cache, billed at the cache write rate, which is usually a premium over the ordinary prompt rate. |
reasoningTokens |
integer | How many of candidateTokens the model spent reasoning, where the provider says. |
totalTokens |
integer | Total tokens the provider reported for the call. |
reportedUsage |
list of object | Every quantity the provider reported, under the provider's own name for it, before anything was added to it or taken from it. |
serviceTier |
string | How the call was served, where a provider prices the same tokens differently by tier, e.g. BATCH, FLEX, PRIORITY. |
cacheWriteTtlSeconds |
integer | How long a cache entry written by this call lives, in seconds. |
region |
string | Where the call was processed, as the client named it, e.g. global, us-central1, us, eu. |
framework |
string | The agent framework or SDK that made the call and its version, e.g. google/adk-go and v1.7.0, as the recorder observed it. |
frameworkVersion |
string | |
providerCostMicros |
int64 as string | What the provider itself said the call cost, in millionths of a US dollar. |
billedCostMicros |
int64 as string | The call's share of what Google billed, in millionths of a US dollar. |
rateCardVersion |
string | The version of the rate card used to price this activity. |
charges |
list of object | Everything else the call was charged for: a per-request search fee, audio or image tokens, a citation token, a grounding query. |
errorClass |
enum | What kind of failure errorCode is, as a value that can be grouped by. |
errorDetail |
string | A short description of the failure, where the product chose to record one, e.g. query must include a date range. |
errorFormat |
string | The convention the failure on this record was reported in, e.g. VERTEX, ANTHROPIC, OPENAI, PERPLEXITY. |
reportedError |
list of object | What the provider said about the failure, under its own names. |
failure |
object | How the failure on this record reads to a person: a label, one sentence, the one value that makes it specific, and everything kept about it. |
observedAs |
enum | Whether the recorder observed this call through an agent framework or from a service's own code: AGENT where it was listening to an agent's callbacks, SERVICE where the call was made directly on an instrumented client. |
project |
string | The unit of work inside a workspace that this call served, e.g. acme-q3-refresh, matter-1183. |
tool |
string | The tool that ran, where this call is a tool call, e.g. web_search. |
attempt |
integer | Which attempt at the same piece of work this call was, counting from one. |
retryOf |
string | The activity this call retried. |
resultBytes |
int64 as string | How much the tool handed back, in bytes and in tokens. |
resultTokens |
integer | How much the tool handed back, in tokens, where something counted them. |
emptyResult |
boolean | The call succeeded and returned nothing the agent could use. |
argsBytes |
int64 as string | How large the tool's arguments were, and a fingerprint of them. |
argsFingerprint |
string | A salted fingerprint of the tool's arguments, where the organisation has turned it on. |
etag |
string | Entity tag. |
createTime |
timestamp | When this activity was created. |
updateTime |
timestamp | When this activity was last updated, which for most activities is when reconciliation wrote billedCostMicros. |
The OpenAPI document has every field. All methods.