Skip to main content
GET
List traces

Authorizations

X-API-Key
string
header
required

API key for authentication

Query Parameters

agent_id
string | null

Return only traces from this agent

conversation_id
string | null

Return only the turns of this conversation. Requires agent_id

Required string length: 1 - 255
q
string | null

Return only traces containing this text in their message ID, conversation ID, conversation history, output, or metadata

output_type
enum<string> | null

Return only traces whose output is of this kind. response covers every trace carrying a reply, including one that also issued tool calls. tool_call covers traces that only issued tool calls

Available options:
response,
tool_call
labels
string[] | null

Return only traces carrying at least one of these labels. Repeat the parameter for each label

Required string length: 1 - 128
score
string[] | null

Return only traces whose latest scoring run matches this condition. Each is written as <evaluator ID>:<condition>, where the condition is passed, failed, or one of =, !=, <, <=, >, >= followed by a number, such as >=4. Repeat the parameter to require every condition at once

created_after
string<date-time> | null

Return only traces created at or after this time (ISO 8601). A time with no timezone is read as UTC. Omit for no earliest date

created_before
string<date-time> | null

Return only traces created before this time (ISO 8601). A time with no timezone is read as UTC. Omit for no latest date

sort_by_evaluator
string | null

Order traces by this evaluator's score on each trace's latest scoring run. Traces this evaluator has not scored come last whichever direction you pick, and an evaluator not linked to the agent counts as having scored none

sort_order
enum<string>
default:desc

Whether sort_by_evaluator puts the highest scores first or the lowest. Ignored without it

Available options:
asc,
desc
include_content
boolean
default:false

Also return each trace's full input, output and metadata. Use limit and offset to page through them

include_score_averages
boolean
default:false

Also return each evaluator's mean score over every matching trace, not just this page. Costs a pass over those traces, so ask for it when the filters change rather than on every page

limit
integer
default:50

Maximum number of items to return

Required range: 1 <= x <= 1000000
offset
integer
default:0

Number of items to skip before returning results

Required range: x >= 0

Response

Successful Response

The page, plus the averages when include_score_averages asked for them.

items
TraceSummary · object[]
required

The page of results

total
integer
required

Total number of items matching the query, before pagination

limit
integer | null

The applied limit (null when unbounded)

offset
integer
default:0

The applied offset

score_averages
TraceScoreAverage · object[] | null

One average for each evaluator linked to the agent that scored at least one trace matching the filters, covering all the matching traces it scored rather than this page. Ask for it with include_score_averages