Most of this list isn't an integration.
Spectre sits on the layer where agents call tools, so anything an agent already reaches — a database, a model, a framework, an API — is visible the moment you point it through. There's nothing to install at either end.
Watched, or wired.
Worth separating, because vendors pad integration pages with things that required no work. Here's which is which.
Nothing to set up
Frameworks, model providers, databases, warehouses, SaaS APIs. Your agent already calls these, and a call is a call — Spectre traces and scores it without knowing or caring what's on the other end.
Which is why this list can keep growing without us shipping anything.
A real integration, built and maintained
Everywhere findings need to go out — your SIEM, your on-call rota, your identity provider, your ticket queue. These have schemas, auth and failure modes we own.
If something you need isn't here, it's a conversation rather than a roadmap request.
What we see in the wild.
Named because customers run them, not because we've certified them. If yours is missing it almost certainly still works — tell us and we'll add it.
Anything marked watched needs no configuration beyond routing your agent's calls through Spectre. Anything marked connector takes credentials and about ten minutes.
Three ways round a missing name.
Nobody should have to wait on a vendor roadmap to get their own data out of a system they're paying for.
Bring the stack you actually have.
Including the parts you'd rather not describe out loud. The awkward ones are usually where the interesting agent behaviour is hiding.