Aetower is designed for local-first performance observation.
Depending on enabled features and macOS permissions, Aetower may observe:
- process names, identifiers, parent/child relationships, bundle paths, and command metadata
- CPU, memory, wakeups, disk, network, GPU-related, thermal, power, and friction signals
- local runtime history and diagnostics
- terminal/session metadata from Chau7 when the Chau7 integration is enabled
- browser tab metadata when a Chromium-compatible debug endpoint is configured
- container metadata when Docker socket access is configured
- local AI-client registration state for MCP discovery
Aetower stores runtime history, diagnostics, settings, and MCP cache data on the local Mac. Developer Preview users should assume these files may contain host and process metadata useful for troubleshooting.
Aetower does not need a cloud service to operate.
Data leaves the machine only when the user explicitly:
- exports JSON or a support bundle
- enables OTLP/HTTP observability export
- shares logs, diagnostics, screenshots, or support artifacts manually
Use the redacted or operator privacy tier for support workflows unless full detail is explicitly required.
Aetower exposes a local MCP surface for supported AI agents. The app-owned live server uses a Unix socket under the user's account. Agents with access to that socket can query Aetower's local observation data.
Only register MCP access for tools you trust.
The default Developer Preview build excludes the optional Endpoint Security helper. If an enterprise build includes it, document that separately because it changes the permission and data-access profile.
- Keep Endpoint Security disabled.
- Keep automatic AI-client registration off unless the user opts in.
- Keep export privacy set to redacted or operator mode.
- Keep telemetry export disabled unless the user configures a collector.
- Keep heavy History and Timeline views in operator-safe mode.