Nothing until you choose.
nano ships no AI provider, key or gateway with nano Self-run. pivt stays off until an admin configures a provider.
pivt writes and tunes detections, turns plain English into nPL, builds parsers from sample lines and helps triage cases. It runs on the provider you configure, including a model inside your own network, and it stays off until you set one up.
pivt works where you already are: the rule editor, the search bar, the parser builder, the case.
| Where | What pivt does |
|---|---|
| Detections | Writes a rule from a description, and proposes tuning for noisy rules. Tuning runs automatically, on request only, or not at all. |
| Search | Turns a question into nPL, reviews and corrects queries, and summarizes results. |
| Parsers | Generates and edits parsers from sample log lines you paste. |
| Dashboards | Builds a dashboard from a description. |
| Cases | Gives triage hints, and can investigate new cases automatically. Automatic investigation is off by default. |
| Automation | Drafts SOAR flows for a person to review and turn on. |
Any OpenAI-compatible endpoint. Works air-gapped.
Not in the path.
nano ships no AI provider, key or gateway with nano Self-run. pivt stays off until an admin configures a provider.
A provider with a base URL is called directly from your install, whether that's a vendor's API or a model on your own hardware.
Summaries send a sample of results, and parser generation sends the lines you paste. If none of that may leave, run the model yourself.
pivt is the AI assistant inside nano. It writes and tunes detections, turns questions into nPL, summarizes results, builds parsers from sample lines, generates dashboards, helps triage cases and drafts SOAR flows. It runs on the AI provider you configure.
No. nano ships no AI provider, key or gateway with nano Self-run. pivt calls only the provider you configure, from your own install, and stays off until you set one up.
Anthropic, OpenAI, Azure OpenAI, AWS Bedrock and Google AI, or any OpenAI-compatible endpoint you run yourself, such as Ollama, vLLM, LM Studio or LiteLLM. Agents that use tools need a model with tool calling turned on.
Yes. Point a provider at an OpenAI-compatible model inside your network through its base URL, and pivt calls it directly. Nothing goes out to the internet.
It depends on the task. Summarizing a search sends the query and a sample of its results. The ai search command sends result rows to the model. Parser generation sends the sample lines you paste. If no log data may leave your network, use a model you host yourself.
No. Give the provider its API base URL, for example the OpenAI API, or the URL of your own model, and requests go straight there. You can route providers through a gateway you run if you prefer.
No. pivt AI is included with nano Self-run at every size. The open-source build does not include it, and nano works fully without AI.
pivt desktop is a separate macOS and Windows workbench for investigations, where an agent's searches open as real tabs and the notes land on the case. It also needs a nano license.
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.