Powered by OpenRouter ZDR01
You ask
Encrypted input
You type your question. It's encrypted in your browser (TLS 1.3) before it ever leaves.
How private inference works
Private by architecture.
Your requests are processed and never stored by AI providers, never profiled. Here's how the zero-data-retention design makes that true.
Anonymous routing
Providers never see you
Process & forget
Nothing stored or profiled
Your private vault
Only you can access
Dual-layer architecture
The brain is split from the memory.
The AI that thinks for you never has access to the memory that stores your history. They're architecturally separated.
Process visualization
How a request flows.
Follow your conversation from question to answer — and see what happens to your data at every step.
02
Anonymous routing
No identity attached
ARMES routes your request without attaching your name, email, or any personal information.
03
Private transit
Secure connection
Your request travels through a private, encrypted connection to the AI provider.
04
AI processes
In-memory only
The AI thinks and generates a response. No persistent storage, no logs, no training data.
05
Immediately forgotten
Process and forget
The AI provider does not retain your request. No record is kept, no profile is built.
06
Your answer arrives
Private storage
The response lands in your private ARMES vault. Only you can access it.
Zero data retention
Nothing is kept after processing.
ARMES routes all AI requests through verified zero-data-retention endpoints. Here's exactly how it works.
How it works
ARMES enforces zero data retention through OpenRouter, routing every request only to endpoints whose providers commit to not retaining your data. OpenRouter tracks each policy at the endpoint level and blocks anything it can't verify.
What isn't kept
The prompt and the response are not retained — no persistent storage, no logs, no training queues, no human-review pools. ZDR endpoints are committed to keeping nothing after processing.
Verification
OpenRouter tracks provider data policies at the endpoint level and, in some cases, negotiates stronger agreements. If a clear ZDR policy can't be established, it conservatively assumes the endpoint retains data and blocks it from routing.
Your data path
Your prompt → encrypted transit → OpenRouter ZDR endpoint → the provider processes it → response returned → the provider keeps nothing → the conversation is stored only in your private ARMES vault. The only persistent copy lives with you.
Privacy guarantees
What private really means.
Not just a promise — four layers of architecture that prevent profiling and monetization by design.
How it works
Process and forget
Every request is routed only to zero-data-retention endpoints — enforced at the account level, not left to each request. AI providers process your request in memory and don't retain it, train on it, or profile you. Your output belongs to you.
Your data, your control
True deletion
When you delete something, it's gone. Not archived, not soft-deleted. No one has it — not even us.
In-transit protection
Encryption in transit
TLS 1.3 encryption on every request, plus AES-256 encryption at rest on stored data. Even if intercepted, your conversations are unreadable noise.
Built for people
Honest engineering
No dark patterns, no hidden data pipelines. ARMES was built because the world needs truly private AI access.
Capabilities enabled by privacy
Private is powerful.
Use the world's best AI for work you'd never type into a tool that profiles you. Private inference isn't just safe — it's freeing.
Deep context memory
Your knowledge base compounds over time — memory that belongs to you and is never shared with AI providers.
Live research
Search and research the web through AI without being tracked, profiled, or targeted.
Cross-agent collaboration
Your agents talk to each other and pull context from one another — transparently, never a black box.
Don't trust promises. Trust architecture.
Verify the zero-data-retention routing yourself, then start using the world's best AI privately.