Say what you believe.
One or two sentences on what the hunt believes and what would rule it out. For example: a macro document that starts encoded PowerShell is a first stage, unless patch management did it.
Detection rules catch what you already know to look for. A hunt tests a hypothesis across your data every time it runs, checks the benign explanations first, and sends a person only the findings that survive two independent AI reviews.
The first three steps are plain queries with no AI. Two AI agents then sort what they found, and a person makes the final call. Later runs learn from it.
Schedule
Due hunts, highest risk first
Discover sources
No data in the window, no AI cost
Run queries
Read-only pivots flag unusual rows
Recheck actors
Known attackers' warning signs
Investigator
Lead, benign or duplicate. Can't escalate
Reviewer
Checks each lead alone. Only step that escalates
Analyst
Acknowledge, open a case or dismiss
| Detection rule | Hunt | |
|---|---|---|
| Question | Did this known pattern happen? | Is this hypothesis true anywhere in my data? |
| Logic | One query with a match condition | A hypothesis, several queries, and AI review |
| Output | An alert for every match | Findings, each labeled lead, benign, duplicate or escalated |
| Benign activity | You add exclusions | Each candidate is tested against a benign explanation, and the hunt remembers what it learns |
| What reaches an analyst | Every alert | Only escalated findings |
| Cost | Query cost | Query cost plus AI tokens on every run |
Hunts and rules work together. A hunt reads the detections that already fired, including quiet ones from rules in Live mode, and labels a finding a duplicate when a detection already covers it. When a hunt keeps finding the same real pattern, turn it into a detection rule. Hunt concepts
One or two sentences on what the hunt believes and what would rule it out. For example: a macro document that starts encoded PowerShell is a first stage, unless patch management did it.
Discovery pivots find candidates. Coverage pivots measure the normal activity around them. Query your events in nPL, or the detections that already fired.
Begin from a template nano provides, or write the playbook yourself in the editor and test each query before it runs. Create a hunt
Hunt queries run as a read-only database user, with limits on time, rows and memory and a required time filter. Hunt security
The investigator can't escalate, only the reviewer can, and AI agents can't dismiss or delete findings. Attacker-written log content is handled as data, not instructions.
Every run uses AI tokens. Set budgets, adjust schedules and pick the model. A sweep with no data in its sources stops before any AI step. Cost and budgets
A hunt tests a hypothesis against your data on a schedule. nano runs its queries, an investigator agent labels each candidate as a lead, benign or a duplicate, and a separate reviewer agent decides which leads to escalate. Only escalated findings reach an analyst.
A detection rule asks whether a known pattern happened and raises an alert for every match. A hunt asks whether a hypothesis is true anywhere in your data, checks the benign explanations first, and sends a person only the findings that hold up. They work together: a hunt reads the detections that already fired, and a pattern a hunt keeps finding can become a detection rule.
No. A hunt never isolates a host, disables an account, closes a case, dismisses a finding or runs a playbook. Its queries are read-only and limited, and the database enforces that too. Escalation can notify your team or open a case for an analyst to acknowledge.
Start from a template that nano provides, or write your own playbook: a hypothesis, up to 12 queries called pivots, a risk weight, a default severity and an escalation mode. Pivots query your events in nPL, or the detections that already fired.
The AI provider you configure in nano under AI & pivt, with a model assigned to each hunt agent: a vendor API with your own key, or a model you run yourself. The investigator and reviewer need models with tool calling. The playbook, query results, what the hunt has learned and the events the reviewer reads go to that provider, and nano doesn't receive your data. Use an on-premises OpenAI-compatible endpoint to keep it all on your network.
Query cost plus AI tokens on every run. You can see what each run cost, set budgets, change the schedule and choose the model. A sweep that finds no data in its sources stops before any AI step, at no AI cost.
Yes, in case mode. When a finding is escalated, nano notifies your team and opens a case that needs an analyst's acknowledgement. Case mode is available only when the hunt passes the case gate; otherwise use notify.
No. Hunts are included with nano Self-run from v2.0.32, at every size. The open-source build does not include them.
Last reviewed 21 November 2026.
From US$149/month at 5 GB/day, US$279/month at 10 GB/day (list price, billed monthly; 10% off billed annually). Every feature in one license, and one license covers two installs.