On-device first
Run Apple Foundation and local model endpoints first. Keep context and tooling fully local whenever possible.
Air-gapped first architecture
Private AI on your phone, with cloud power only when you say so.
Aegis is an iOS AI terminal built for people who care about data boundaries. Use on-device and local LLMs by default, attach private tools, and selectively delegate tasks to cloud LLMs through strict redaction rules.
No automatic telemetry
No automatic diagnostics upload
User-initiated debug sharing only
On-device tools: calendar, reminders
Per-model tool controls
Import/export full settings JSON
Built for privacy-critical workflows
Aegis keeps local conversations and tool execution on your device. Cloud models are optional and can be gated with strict redaction before any request leaves your phone.
Strict cloud delegation
When cloud delegation is enabled, Aegis can apply redaction policies before sending prompts to external providers.
Tool-gated execution
Enable or disable each tool globally and per source/model. Prevent accidental tool execution by default.
What Aegis does not do by default
The trust boundary is explicit: local unless you intentionally export or send.
Default behavior
- No account requirement
- No automatic diagnostics upload
- No automatic conversation sync to remote servers
- No hidden background analytics pipeline
User-controlled actions
- Manual debug share (you choose what to export)
- Per-source cloud model enablement
- Per-tool enable/disable controls
- Import/export full settings and model profiles
Use cases
Security teams, founders, researchers, and consultants use Aegis where privacy and traceability matter.
Personal assistant with local tools
Manage calendar and reminders through explicit tool calls visible in the conversation timeline.
Hybrid model strategy
Route routine requests locally, and escalate only selected tasks to cloud providers after redaction.
Auditable debugging
Export full diagnostics only when needed for support. No silent background diagnostic streams.
Aegis TestFlight link on this page is a mock placeholder for now.