Open-core engine in Rust. Sub-second search on ClickHouse, in one store and one language. Everything below is in nano and runs inside your environment; sections marked Open source are also in the free AGPL edition.
Rust at the core, ClickHouse for storage, Vector for ingest. No JVM warmups, no garbage-collection pauses, no surprise memory limits. The whole stack picked for one job, which is getting your team to an answer.
The core engine, the detection rules, the parsers and the docs are in the open: the engine is AGPL-3.0, free to audit, fork and self-host. The paid license adds the commercial modules (cases, pivt AI, entity risk analytics, SSO, HA, Kubernetes, air-gapped operation) on the same installer. Two editions, one installer.
Pipe-based, left-to-right, familiar if you've used a search-and-pipe SIEM. Stricter, faster, and the autocomplete actually knows your schema. It works on any logs, so ops teams use it for log management too.
1class_uid=3002 activity_id=1 status_id=2
2| stats count as attempts by src_endpoint.ip, user.name
3| where attempts > 10
4| sort -attempts
5... 1class_uid=4002 earliest=-7d
2| prevalence enrich=true window=7d
3| where host_count < 3 AND is_rare=true
4| table http_request.url.hostname, first_seen, host_count
5... 1class_uid=1007
2| prevalence enrich=true
3| tree process root="*powershell*"
4... 1class_uid=4002
2| bin span=10m
3| stats count by time_bucket, http_request.url.hostname
4| stats stdev(count) as jitter, avg(count) as avg_hits by http_request.url.hostname
5... YAML frontmatter for metadata, pipe-language for the actual hunt. MITRE mapping, per-rule risk scores and prevalence enrichment are first-class, not glued on with a script.
1---
2title: regular_beacon_interval_detection
3severity: high
4mode: staging
5schedule: "*/10 * * * *"
6mitre_tactics: TA0011
7mitre_techniques: T1095
8---
9class_uid=4002
10| where NOT cidr_match("10.0.0.0/8", dst_endpoint.ip) AND NOT cidr_match("192.168.0.0/16", dst_endpoint.ip)
11| stats count, dc(time) as unique_times by src_endpoint.ip, http_request.url.hostname
12| where count > 20 AND unique_times > 15
13| prevalence enrich=true window=30d
14| risk score=70 entity=src_endpoint.ip factor="Regular beacon pattern"
15| table src_endpoint.ip, http_request.url.hostname, count, unique_times, host_count, risk_score Cases bundle the alert, the evidence, the timeline, and the analyst's notes in one place. Notebooks for ad-hoc hunting, and scheduled hunts that send you only the findings that hold up. Process trees and identity resolution for the messy parts. And SOAR flows to contain, suspend and ticket, with a person approving the disruptive steps.
pivt, nano's AI assistant, triages alerts, drafts parsers, follows playbooks and writes case notes. Configurable, auditable, and never the one closing the ticket without you.
Off by default. Point pivt at any OpenAI-compatible endpoint, including self-hosted models, or the Anthropic Messages API with your own key. Model calls go straight to that endpoint, never through nano. More about pivt AI
Define steps in YAML. pivt runs them on the alerts you choose: pull EDR context, check threat intel, query identity, draft a summary. Stop a step for human approval whenever you want.
Drop a sample log, get a working parser. pivt reads the format, drafts the parser, validates it against your data. The 80% case in two minutes, not two days.
Write incident response runbooks once. pivt executes them on the right alerts, asking for human confirmation on the steps that need it. Auditable trail of every action.
By default, events are enriched on your servers at ingest from one signed feed bundle: IPinfo Lite (IP country, continent and network owner), ThreatFox indicators and the Tor exit-node list. The bundle updates hourly; turn it off or mirror it. Everything else, from asset context to lookups with your own API keys, is yours to configure. What the feeds send (nothing)
Every install starts with zero log sources. The catalog of parsers, detection rules and enrichments syncs from nano as signed, versioned content, and you choose what to onboard. Or fork them and make them yours. Already run OpenTelemetry? Send it over OTLP.
Traffic, threat, config, and system logs from PAN-OS 10+ firewalls.
EDR detections and telemetry, in both FDR and Streaming API formats.
System Log events: authentication, user lifecycle, MFA, and admin actions.
API calls, user identity, and resource changes across your AWS account.
EDR process, network, file, and logon events.
Network telemetry: conn, dns, http, ssl, files, ssh, and more.
The catalog lives inside nano. Browse and choose without leaving the product. Every artifact is open source, so fork and contribute through GitHub.
Risk scores stack across signals. Rules contribute weighted factors, prevalence baselines tell you what's actually rare, and the score updates in real time as more events land.
Dashboards that compose. A health page that watches the pipeline, with pivt's health analysis every 12 hours once you configure a model. And settings deep enough to make every team's nano feel custom, without a fork.
Five steps. No services engagement. No "professional services" line item on the invoice. Most teams hit step 5 the same afternoon.
Full walkthrough in the install guide, or let your coding agent do it with the nano-deploy skill. Deployment options in the nano overview.
Buy nano online, run the demo, or install Open source and ingest something this afternoon.
Open source free under the AGPL · One license, every feature at every size · Read the docs