# Hunt Query Builder

Build a hunting query in KQL, SPL, LogScale or PowerQuery against documented field names, or start from a prebuilt hunt.

Build a hunting query in the language your platform speaks: KQL, SPL, LogScale or PowerQuery. Every field name comes from that platform's own documentation, and the prebuilt hunts load straight into the form.

## How to read it

Start from a prebuilt hunt when one fits. Each says what it finds, what data it needs and where the noise comes from.

Filter as early as you can, aggregate after, and sort before you truncate. Those three habits fix most slow or wrong hunting queries.

Field names do not map one to one between platforms, so each tab builds its own query rather than translating another tab's.

## Tab by tab

### KQL

Open it: https://www.cyberkit.win/tools/hunt-query/?tab=kql

Build a Kusto query for Defender XDR advanced hunting or a Sentinel workspace, one clause at a time. It writes query text and nothing else. There is no tenant connection, so you copy the result and run it yourself.

Start from a hunt, then edit it. Each one says what it finds and how to tune it before you trust the results.

Use the case-insensitive operators (=~, in~, has) for file names and command lines. Attackers change case, and == is case-sensitive.

It covers one table piped through where, summarize, project, sort and take. Joins and let statements are out of scope.

### SPL

Open it: https://www.cyberkit.win/tools/hunt-query/?tab=spl

Build a Splunk search as a base search plus a pipeline of commands. It writes SPL. It does not run it.

Always set a time range. A triage search without one is an all-time scan of every index you can read.

Order matters more than syntax. Filter before stats, because once you aggregate, the field you wanted to filter on is gone.

Stats takes aggregations as count, dc(user) as users. Sort takes fields with a leading minus for descending.

### LogScale

Open it: https://www.cyberkit.win/tools/hunt-query/?tab=logscale

Build a Falcon LogScale query for Advanced Event Search. It writes query text with no connection to any tenant.

Pick the event first. ProcessRollup2 is where nearly every hunt starts: what ran, as whom, from where.

Check field names against the Event Data Dictionary in your own console. A field name that is slightly wrong returns zero rows, and zero rows looks like nothing to find.

Use groupBy to stack-count. The rare values at the bottom are usually the interesting ones.

### PowerQuery

Open it: https://www.cyberkit.win/tools/hunt-query/?tab=powerquery

Build a SentinelOne PowerQuery for Singularity Event Search: a filter expression, then group, filter, sort, limit and columns.

Terms in the filter are joined by a space, which means AND. A query grows by adding conditions, not by nesting them.

src.* is the parent process and tgt.* is the child. Mixing them up is the most common reason a hunt returns nothing.

Use contains:anycase for names and command lines. Plain contains is case-sensitive.

## Related

- [KQL Cheat Sheet](https://www.cyberkit.win/reference/kql-cheatsheet/)
- [SPL Cheat Sheet](https://www.cyberkit.win/reference/spl-cheatsheet/)
- [Rule Range: Query pipeline](https://www.cyberkit.win/drills/rule-range/?stage=pipeline)

Canonical: https://www.cyberkit.win/tools/hunt-query/

This runs in the browser. Nothing you enter is uploaded.
