Experiments API
- HobbyAvailable
- CoreAvailable
- ProAvailable
- EnterpriseAvailable
- Self HostedLangfuse v4+
The Experiments API lets you retrieve experiment data from Langfuse for analysis, evaluation pipelines, notebooks, and CI/CD workflows. An experiment is a run of your application against test data. Each experiment item represents one input, its expected output, and the actual output produced by your application.
For the complete request and response contract, see the Experiments API reference.
Choose the right API
| If you want to... | Use |
|---|---|
| List experiment runs and their summaries | GET /api/public/experiments?fromStartTime=2026-01-01T00:00:00Z |
| Retrieve experiment items and their inputs, outputs, expected outputs, metadata, and scores | GET /api/public/experiment-items?fromStartTime=2026-01-01T00:00:00Z |
| Retrieve the complete trace and observation tree for an item | Observations API v2, using the item's traceId |
| Query evaluation scores independently | Scores API v3 |
The experiment endpoints support filtering, cursor-based pagination, and optional response fields. Refer to the Experiments API reference for the available filters and response fields.
Experiment-level and item-level scores
Experiment-level scores summarize the complete run. Item-level and trace-level scores evaluate individual experiment items. The Experiments API returns both levels in their corresponding responses, while the Scores API v3 is useful when scores are the primary data you want to query.
Related resources
- Experiments via SDK — run experiments with the Python or JS/TS SDK.
- Experiments data model — understand how datasets, experiments, items, traces, observations, and scores relate.
- Observations API v2 — retrieve row-level trace and observation data.
- Scores API v3 — retrieve evaluation and annotation scores.
Last edited
