Tip of the spear.

For a single team, use it. For a service provider it has no native multi-tenancy, and organisation segregation is an Enterprise Edition feature; one instance per client is exactly the overhead you are trying to escape. And it has no sensors of its own.

Keep MISP. We interoperate with it: MISP feeds in, a MISP feed out, and sharing groups that push and pull over the MISP REST API. But hosting it properly is an infrastructure project and it ships a dozen or more upgrades a year. We are the operating layer above the standard, not a replacement for it.

Take it. It is excellent. It gives you daily reports about your own netblocks. It does not correlate them with your other sources, tell you which exposure is being probed on your own sensors this week, produce enforcement-format output or a branded deliverable. Attack Surface Management does the outside-in view from seeds you prove you own.

Then choose. Run it on-prem (Docker Compose on a VM, or Helm for high availability), take a tenant with its own database on Wahukeza-operated instances, or combine the two. Models can be changed later, with export in STIX 2.1, CSV and JSON. On-prem ships with a migration runner, signed images with SBOMs, and upgrade and restore drills as release gates.

Low-interaction ones do. Ours are high-interaction personas: cloned management UIs and command lines, an Ivanti Connect Secure gateway, SAP NetWeaver, a decoy MCP server, a camera that streams. The device clones are replicas of real devices built by a capture engine that records the management surface of the kit you actually run, not hand-written fakes. Exploit watch maps sessions to a cited catalogue of named exploitation campaigns, so you see which ones are aimed at the kit you run.

Attacker-facing code cannot execute or dial out: static tests forbid exec, subprocess, eval and any outbound client in the transports, the login wall, the web, lure and TDS modules and the Dahua, MikroTik, Ivanti, MCP and SAP modules, and each CLI engine has its own security suite. Every outbound verb (wget, curl) is answered, never performed. UDP is never served, so a sensor cannot be a reflector, and uploaded bodies are stored inert.

Four packages (Basic, Standard, Advanced, Complete) and every module except the two AI modules can be licensed on its own. Each is an annual licence per installation or tenant with unlimited analysts. Pricing follows a short scoping workshop, because sensors, endpoints and delivery model are what move it.