Edge-first architecture

The interaction starts before the cloud sees it.

AxLoop begins where users and agents initiate work, preserves privacy-controlled evidence on the device, and provides the foundation for reconstructing supported AI interactions.

Talk to AxLoop

Available now / passive evidence pipeline

The evidence foundation for interaction observability.

AxLoop observes supported surfaces on the device, classifies evidence locally, and can export normalized inventory through OpenTelemetry. Correlation and tracing build on this foundation.

01
Edge evidenceBounded, allowlist-based discovery
device boundary
Apps + processesCatalog-matched software
running
CLI toolsSupported local tooling
configured
IDE extensionsVS Code + JetBrains
configured
MCP declarationsSupported config sources
configured
Models + runtimesLocal runtime metadata
running
Local containersSupported runtimes
running

Privacy boundary No prompts, responses, source code, credentials, raw paths, or tool payloads.

02
AxLoop local serviceNative collection + durable state
1
DiscoverPeriodic reconciliation + approved config events
2
Classify + redactRisk categories and privacy controls run locally
3SQLiteInventory state
4OutboxDurable retry queue
5
NormalizeSchema v2 evidence events
Processing and storage stay local by default
03
Your environmentCustomer-controlled destination
opt-in
OTLP collectorLogs over HTTP / protobuf
Fleet aggregationCorrelate evidence across devices
Your operations stackSIEM, data lake, or observability platform
Export contractaxloop.asset.observedaxloop.device.inventory
At-least-once delivery with state-change deduplication
Evidence statesConfigured declarationRunning evidenceVerified connection**Not inferred from passive discovery. Requires separate host integration or an explicit user-initiated probe.
01

Edge collector · Available now

The native macOS service observes supported processes, applications, extensions, model metadata, containers, and MCP configuration sources without arbitrary filesystem crawling.

02

Local evidence pipeline · Available now

Classification, privacy controls, inventory state, and a durable delivery outbox remain on the device. Export is off by default.

03

Customer-controlled telemetry · Available now

Optional OTLP export sends normalized evidence to infrastructure you control, supporting fleet reconciliation without exporting prompts, source code, or tool payloads.

04

Interaction correlation · Near term

AxLoop is extending discovery evidence to correlate supported process, parent-process, connection, endpoint, agent, model, MCP, and timestamp relationships.

05

Trace and evaluation · Product direction

The platform direction is to reconstruct the observable sequence from agent to tool to application, then compare expected and observed outcomes.

06

Policy, enforcement, and audit · Product direction

Governance will build on trustworthy traces, with policy decisions, human approval, enforcement at supported control points, and a defensible event history.

AxLoop AI

Build the interaction layer from evidence.

Discover first. Correlate next. Never present inference as observation.

Book a working session