Local-first, not local-optional
Telemetry stays inside your perimeter by default. Nothing about an agent's behaviour is routed through us to make the product work.
Company
Observability that belongs to you.
AxLoop AI is building the telemetry layer for AI agents wherever they actually run — on devices, not just servers. It's a collector, context-propagation and federation problem, and we're treating it like one.
01
Principles
How we build it.
Open standards over lock-in
Spans are plain OpenTelemetry with W3C TraceContext. If you want to send them somewhere else tomorrow, you can.
Redact at the edge
Sensitive parameters are redacted before anything is written to disk, so the safest place for data is where it already is.
Useful before it is paid
The collector is open source and works on its own. The fleet layer earns its keep only once you actually run a fleet.
02
Who it's for
Teams whose agents already left the datacenter.
Engineering & platform teams
Teams running AI agents across laptops, phones and IDEs that need latency, error and cost visibility they can act on.
Security & governance
Anyone who has to answer which tools ran, on which device, and what data passed through them.
Regulated environments
Organizations that cannot send agent telemetry to a third-party SaaS and need a deployment that lives in their own infrastructure.
Early access
Bring us your real agent fleet.
AxLoop AI is early. We're onboarding a small number of teams running agents across real devices, and shaping the collector around what they find.
If that sounds like your fleet, get in touch with the team to join the early-access list.
Request early access