Investigate your network in plain English: introducing Natural Language Query

Every SOC analyst has been there. An alert fires. You know what you need to find. Maybe it's all outbound connections from a specific host that spiked overnight. Maybe it's every DNS query over 100 characters from a subnet you're watching. You know the question. What you don't know is the exact query syntax you need to ask it.

So you open the documentation. You search for field names. You try a query, get it wrong, adjust, and try again. Minutes pass. During an active investigation, those minutes are expensive.

This is the query syntax barrier, and it affects every analyst who’s not a seasoned expert in the query language. It's not a security knowledge gap. It's a gap between knowing what to look for and being able to pull it from the data.

With the v29.2 release, we're closing that gap.

Ask the question, get the answer with Natural Language Query

Natural Language Query (NLQ) translates plain-English questions into working LogScale Query Language (LQL) queries and executes them against your data. You do not need syntax memorization or documentation tabs. You type what you want to know, and the NLQ agent handles the translation.

For example:

“Show me all connection spikes on port 443 last night.”

“Which internal hosts communicated with IP addresses in countries we've never seen before?”

“How much data did 10.4.12.88 upload in the last 48 hours?”

Each of these becomes a functional LQL query that executes immediately, with results returned in Investigator. The analyst stays focused on the investigation instead of fighting with syntax.

Why this matters now

Natural language security queries aren’t new in the SIEM or NDR space. Several products have offered different flavors of NLQ. Corelight is taking the NLQ concept into the agentic future while staying true to its meaning.

What makes NLQ really powerful here is the evidence Corelight provides. Corelight generates the richest network telemetry in the industry. It includes connection records, protocol metadata, threat intelligence, and behavioral and machine learning detections. The depth of evidence is extraordinary. The problem has always been that accessing that depth required deep knowledge of the telemetry and specialized query skills that most analysts simply don't have time to build.

NLQ unlocks the full power of that evidence for every analyst on the team, on day one, regardless of their knowledge of Zeek or fluency with LQL.

What it looks like in practice

Consider a common investigation scenario. An L1 analyst receives an alert about a potential DNS tunneling detection. Before NLQ, the investigation path looked something like this:

1. Open the alert in Investigator
2. Search documentation for the right dns.log fields
3. Attempt a query filtering on query length and entropy
4. Realize the syntax was wrong, adjust, and re-run
5. Get results, then pivot to the source host
6. Write another query from scratch for the conn.log data

With NLQ, that same analyst types: “Show me all DNS queries from 10.4.12.88 where query length exceeds 50 characters in the last hour.” Results appear. Next query: “Show all outbound connections from 10.4.12.88 over the last 24 hours sorted by bytes transferred.” Another set of results. The investigation keeps moving.

The time savings compound quickly. Our internal testing showed that analysts who previously spent 3 to 5 minutes constructing a single query could get results in seconds. Across a full investigation with multiple pivots, that can be the difference between a 30-minute triage and a 5-minute one.

Paired with the Corelight Agent Builder Library

NLQ becomes even more valuable when paired with the Corelight Agent Builder Library, which we're also shipping in this release. The Agent Builder Library provides structured investigation playbooks that tell analysts (and their AI agents) what to investigate for any given alert type. NLQ removes the barrier to actually pulling the evidence those playbooks call for.

Think of it this way: The playbook says “check for anomalous upload volume from the source host.” NLQ lets any analyst find that data immediately, without needing to know that the right approach is a specific groupBy aggregation over conn.log with a bytes filter on orig_bytes.

The combination means analysts know what to look for and can find it instantly. That's a meaningful change for L1 teams that previously had to escalate network-heavy investigations to senior staff.

How it works

NLQ translates natural language into LogScale Query Language and executes the resulting query directly against your Investigator data. The translation is context-aware. It understands Corelight's log schema, field names, and common investigation patterns. When you ask about “connections,” it knows you mean conn.log. When you ask about “DNS queries to unusual domains,” it knows which fields are relevant and how to filter them.

A few things worth noting about how we built this:

• It executes, not just translates. You don't get a query string that you then have to copy and paste into a search bar. NLQ runs the query and returns results directly. The generated LQL is visible, so you can learn from it, modify it, or save it for reuse. But the default path is: Ask, receive, act.
• It works with your data as it exists. NLQ queries run against whatever logs your sensors ship to Investigator. There's no separate index, no preprocessing required. If the data is in Investigator, NLQ can query it.
• It shows its reasoning. Every generated query comes with a collapsible analysis panel that explains how NLQ interpreted your question: What logs it identified as relevant, what criteria it extracted, and how it constructed each part of the query. You can expand this to learn LQL over time, or collapse it and move on. Either way, the logic is transparent.

Who benefits most

Every Investigator customer gets NLQ at GA, once they enable the GPT private data integration. A few groups will feel the impact the most:

• L1 and L2 analysts: They can now investigate network evidence directly instead of escalating alerts they don't have the query skills to pursue.
• SOC leads and managers: NLQ makes it easier to build a high-quality query library. Users describe what they need in plain language, receive a polished query, and can save it for the whole team to reuse in dashboards and workflows.
• Incident responders: They need fast answers during active triage. When you're confirming breach scope at 2 AM, the last thing you want is to debug query syntax.
• Teams building custom dashboards and visualizations: They no longer need specialized LQL knowledge to create new panels or widgets.

Getting started

NLQ is available at GA to all Investigator customers who enable the GPT private data integration. You'll find it in the Investigator search interface, ready to use.

For teams evaluating Corelight, NLQ is part of a broader investment we're making to remove barriers between analysts and network evidence. The Corelight Agent Builder Library provides the investigation expertise. NLQ provides the query access. Together with Agentic Triage verdict export (which delivers pre-investigated findings directly to your SIEM), the entire investigation workflow becomes faster without requiring new specialized skills.

If you're a current customer, enable the GPT Private Data integration and start asking questions of your network data in plain English. If you're evaluating Corelight, reach out for a demo to see NLQ in action against live network data.