Query with Explore

Run a query

Open Explore in the sidebar. Choose a time range, then enter a query in the Query tab. Press Enter or select Run. Explore works without a configured AI provider.

Count spans by service:

spans | count() by service

Plot p95 span duration by service in five-minute buckets:

spans | p95(duration) by service every 5m

Find warning and error logs from checkout:

logs where service:checkout severity>=warn

Use service names present in your telemetry. See the FanQL reference for syntax, fields and functions.

Use the Builder tab

Select Builder to choose a source, measures, filters, grouping fields, time buckets and a row limit. Query and Builder edit the same FanQL query. Builder changes appear under As FanQL. Select Run to execute. Complex filters and measures remain as chips; edit them in Query. If the Query tab says the text is read-only, keep editing in Builder.

Use the SQL tab

Select SQL to write a separate statement. Switching tabs does not convert FanQL to SQL. Run one read-only SELECT with a time filter, for example:

SELECT service, count(*) AS spans
FROM spans
WHERE $__window(start_time)
GROUP BY service

Explore shows SQL results as a table. See the guarded SQL rules and rollup relations for limits and rollup columns. The get_telemetry_schema MCP tool lists the columns of spans, logs and metrics.

Read diagnostics

When a query has problems, the editor marks the affected text and lists messages below it. Problems prevent execution. Use Replace with when a suggested fix fits, or edit the text yourself, then run again. Notices provide guidance and do not block execution.

The result’s Ran as line shows the canonical query. If you edit a query or a run fails, the previous result can remain visible. Check its status before reading it as the result of your current query.

Pick a visualization

After parsing, the visualization picker offers types compatible with the query. Choose one to rerun it. Options that need settings such as minimum and maximum values or thresholds stay disabled until those settings are supplied in a dashboard panel.

Buttons with an ellipsis offer a query rewrite. Hover to inspect the proposed text before selecting one. The rewritten query appears in the editor and runs.

Search Logs and Traces

Open Logs or Traces in the sidebar. Their search bars accept FanQL filters only. Leave out the source, where and aggregation clauses.

On Logs, enter:

service:checkout severity>=warn

On Traces, enter:

status:error duration>500ms

Press Enter or select Search. The service, severity, error and duration controls add filters to the typed search. Use the time picker and order control to change the window and ordering.

Trace results list traces with matching spans. The displayed root span can have different values from the span that matched; see trace semantics.

Save to a dashboard

After a successful run with data, select Save to dashboard in Explore, Logs or Traces. If you changed the query or filters, run them again first.

Choose an existing dashboard or New dashboard, enter a panel title and select Save. A new dashboard also needs a name. Pitot opens the saved dashboard. Existing dashboards keep their time settings; new ones use the selected window. See dashboard controls for further changes.

Use HTTP or MCP

The HTTP API reference lists FanQL parse, format, completion and reference routes, plus panel query and dashboard routes. Over MCP, query_telemetry accepts FanQL text with a panel. Use create_dashboard or edit_dashboard to save panels.