Evals | 514 Docs

Route	Use	CLI equivalent
`GET /evals`	List evals.	`ax evals list`
`GET /evals/{eval_id}`	Get an eval.	`ax evals show`
`GET /evals/{eval_id}/trend`	Read grader trends.	`ax evals trend`
`GET /evals/{eval_id}/results`	List eval results.	`ax evals results`
`GET /evals/{eval_id}/results/{run_id}/{case_key}`	Read a case result.	`MCP: eval_result_get`
`GET /evals/{eval_id}/results/{run_id}/{case_key}/detail`	Read result assertions and events.	`MCP: eval_result_detail`
`GET /evals/{eval_id}/runs`	List eval runs.	`MCP: eval_runs`
`GET /evals/runs/{run_id}`	Read an eval run.	`MCP: eval_run_get`

All routes require a bearer API key and organization access with agent evals enabled. Supply org_id when the key belongs to several organizations. GET parameters use snake_case. Write envelope fields use lowerCamelCase; fields inside values keep snake_case (case_key, split_ratio). Records retain their original snake_case fields. Arguments are optional unless marked required. List routes return {rows, nextPageToken}. Pass page_size (default 50, maximum 200) and page_token to continue. Single-record routes return {record}.

Test-split results and snapshot cases stay redacted until your user explicitly records a reveal for that exact dataset version. The response preserves redacted_held_out: true. Detail reads return HTTP 412 with error: "failed_precondition" and a message naming RecordReveal when a reveal is required. No GET records a reveal.

`GET /api/v1/evals`

List evals.

curl -H "Authorization: Bearer $AX_API_KEY" \
  "https://app.514.ax/api/v1/evals?agent_id=$AGENT_ID"

Parameter	What it does
`agent_id`	Required. Context agent UUID.

`GET /api/v1/evals/{eval_id}`

Get an eval.

curl -H "Authorization: Bearer $AX_API_KEY" \
  "https://app.514.ax/api/v1/evals/${EVAL_ID}?agent_id=$AGENT_ID"

Parameter	What it does
`agent_id`	Required. Context agent UUID.

`GET /api/v1/evals/{eval_id}/trend`

Read grader trends.

curl -H "Authorization: Bearer $AX_API_KEY" \
  "https://app.514.ax/api/v1/evals/${EVAL_ID}/trend?agent_id=$AGENT_ID"

Parameter	What it does
`agent_id`	Required. Context agent UUID.

`GET /api/v1/evals/{eval_id}/results`

List eval results.

curl -H "Authorization: Bearer $AX_API_KEY" \
  "https://app.514.ax/api/v1/evals/${EVAL_ID}/results?agent_id=$AGENT_ID"

Parameter	What it does
`agent_id`	Required. Context agent UUID.
`split`	Optional train, validation or test filter.

`GET /api/v1/evals/{eval_id}/results/{run_id}/{case_key}`

Read a case result.

curl -H "Authorization: Bearer $AX_API_KEY" \
  "https://app.514.ax/api/v1/evals/${EVAL_ID}/results/${RUN_ID}/${CASE_KEY}?agent_id=$AGENT_ID&dataset_id=$DATASET_ID"

Parameter	What it does
`agent_id`	Required. Context agent UUID.
`dataset_id`	Required. Parent dataset UUID.

`GET /api/v1/evals/{eval_id}/results/{run_id}/{case_key}/detail`

Read result assertions and events.

curl -H "Authorization: Bearer $AX_API_KEY" \
  "https://app.514.ax/api/v1/evals/${EVAL_ID}/results/${RUN_ID}/${CASE_KEY}/detail?agent_id=$AGENT_ID&dataset_id=$DATASET_ID"

Parameter	What it does
`agent_id`	Required. Context agent UUID.
`dataset_id`	Required. Parent dataset UUID.
`assertions_page_token`	Optional assertions cursor.
`events_page_token`	Optional events cursor.

Returns {result, assertions, events, assertionsNextPageToken, eventsNextPageToken}. Each evidence list has an independent cursor; page_size bounds each.

`GET /api/v1/evals/{eval_id}/runs`

List eval runs.

curl -H "Authorization: Bearer $AX_API_KEY" \
  "https://app.514.ax/api/v1/evals/${EVAL_ID}/runs?agent_id=$AGENT_ID"

Parameter	What it does
`agent_id`	Required. Context agent UUID.

`GET /api/v1/evals/runs/{run_id}`

Read an eval run.

curl -H "Authorization: Bearer $AX_API_KEY" \
  "https://app.514.ax/api/v1/evals/runs/${RUN_ID}?agent_id=$AGENT_ID"

Parameter	What it does
`agent_id`	Required. Context agent UUID.

Search documentation