Edge clients
A Rust desktop app, a Swift mobile app, a Python CLI, a Node IDE extension — whatever mix you run. Each emits telemetry differently — or not at all. The collector normalizes all of them into one OTel schema.
Architecture
Telemetry that never leaves your perimeter.
Aggregating hundreds of simultaneous agent devices into a coherent fleet view, with local data residency, is a hard distributed systems problem. Here's how AxLoop AI approaches it.
PHONE APP
CLAUDE DESKTOP
DEV LAPTOP
BROWSER AGENT
↓ AXLOOP AGENT · SDK / DAEMON / OS HOOK ↓
LOCAL OTEL COLLECTOR · PER DEVICE
ENCRYPTED · BATCHED SYNC ↓
FEDERATED AGGREGATION NODE · ON-PREM / VPCAXLOOP PLATFORM · traces · logs · metrics · per-device · fleet view
OpenTelemetry nativeW3C TraceContextRust · Swift · Python · Node · GoClickHouse · Tempo · LokiKubernetes or Docker Compose
AxLoop AI agent
SDK, daemon, or OS hook depending on the surface. Applies redaction policy before anything is written to disk, so sensitive parameters never persist.
Local OTel collector
Per device, language-agnostic. Buffers locally when offline, then ships encrypted batches. The device stays useful even when the network isn't.
Federated aggregation node
Runs on-prem or in your VPC. Kubernetes for real fleets, single Docker Compose for smaller deployments. Reconciles spans from hundreds of devices arriving out of order.
AxLoop AI platform
Traces, logs and metrics with per-device and fleet-wide views, RBAC so teams see only their own agents, and a full audit log with conversation replay.
Key technical choices
The decisions behind the design.
OTel-native from day one
No vendor lock-in. Point your spans elsewhere tomorrow if you want to — which is exactly why teams trust the collector enough to install it.
W3C TraceContext propagation
Spans link across process and network boundaries, so an agent's reasoning step connects to the tool it called and the downstream API it hit.
Storage that speaks the ecosystem
ClickHouse for metrics, Tempo-compatible traces, Loki-compatible logs. Plug into the Grafana you already run.
Local data residency by default
Teams in fintech, healthcare and the public sector often cannot send agent telemetry to a SaaS. Deployment location is a configuration, not a special build.
Data residency
Local-first deployment is the default here, not a special build.
Many organizations simply cannot send agent telemetry to a third-party SaaS.