Every capability ships in the open-source Open tier, Iris included. Nothing is gated. The paid tiers sell the work around it: managed hosting, support, and bespoke install and integration.
The entire platform. You run it, you own it, you keep every byte. Bring your own model or run Ollama locally.
Deploy on GitHubWe host, patch, and keep it running. Same platform as Open, none of the ops. Redaction on by default before any LLM call.
Get started →We install it on your perimeter, integrate it with your stack, and build what your org specifically needs. SSO, SLA, 24/7 support.
Talk to us →Yes. The Open tier uses the same Docker images as Pro and Sovereign. Migration is a config change, not a re-deployment.
Your choice, on every tier. Run Ollama or vLLM locally, point it at any OpenAI-compatible API, or disable LLM calls entirely and stay 100% deterministic. The managed Pro tier defaults to Claude Sonnet with redaction on.
No. SSVC prioritization, CVE correlation, KEV matching, EPSS scoring, and Watchtower alerts are 100% deterministic. Zero LLM calls, zero per-query cost.
Before any prompt reaches the LLM, Horus replaces hostnames, IPs, and emails with stable per-org pseudonyms (host-a, ip-1, user-x), then de-pseudonymizes on the response side. Your actual infrastructure identifiers never appear in an LLM prompt. On the Open tier, point Horus at a local Ollama instance and nothing leaves your network at all.
On the managed Pro tier the LLM is Claude Sonnet with redaction on by default. On Open, swap it for any endpoint or run Ollama locally for zero egress.
Yes. The live demo is pre-loaded with 30 days of scan history, real CVE findings, Red/Blue debate transcripts, and phishing campaign results. No card required.
On-prem or air-gapped installation, integration with your existing stack, and development of the specific capabilities your org needs. Plus SSO/SAML, a dedicated SLA, and 24/7 priority support. Scoped and priced per project.
The demo is pre-loaded with 30 days of posture history, real CVE findings, Red/Blue debate transcripts, and phishing campaign results.