Privacy is not a feature. It's a right.
Zero-knowledge encrypted messaging that keeps working when the network doesn't — one audited Rust crypto core, five native apps, and a server that can't read a thing.
windows · macos · linux · ios · android — shipped at v0.16, audit-track next
Even the "secure" ones leak more than they admit. Seven ways the incumbents fall short — and the reason VAULTEX was built to close each one.
Metadata collectionServers know who talks to whom, when, and how often.
Phone number requirementsLinks your real-world identity to every message.
Server-side contact listsYour social graph is stored and analyzable.
Centralized trustYou must trust the provider not to read your messages.
Backdoor pressureProviders face legal compulsion to comply, quietly.
Closed-source cryptoYou can't verify the encryption actually works.
IP address loggingYour location is tracked with every connection.
Stores only encrypted blobs. Cannot read messages, see contacts, or identify users.
No phone number or email. Purely cryptographic identity. You control your keys.
Server cannot determine who sent any message. Social graph is invisible.
The crypto crate is public today. No proprietary black boxes in the crypto path. github.com/VaultexCoder/vaultex →
Backdoor mandates and metadata subpoenas are now routine policy fights. Buyers in public safety, healthcare, and legal want communications they can audit, not promises they have to believe.
Disaster response and field operations need messaging that survives the network going down. Mesh relay and BLE on commodity phones are finally practical — and VAULTEX already ships them.
A five-platform product with audited cryptography used to need a large team. VAULTEX was built by one founder with AI agents under a security-gated process. That window is open now, and it won't stay open.
VAULTEX was built end to end by a single engineer, paired with AI coding agents under a security-gated process — and that engineer has spent a career building communications that had to hold up under adversaries, auditors, and accreditation.
25+ years developing systems that run on classified networks: communications, satellite RF, electronic warfare, PNT, radar, hypersonics, and missile defense.
NSA-approved encryption tooling; SIPRNet and NIPRNet transport. Designed satellite and ground-control systems on a bifurcated architecture — classified and unclassified pathways, cross-domain solutions, encrypted tunnels. Zero-knowledge isn't a slogan here; it's the operating assumption the founder has built under for decades.
An RF engineer built the radio side. LAN, Bluetooth LE, WiFi Direct, and mesh relay weren't bolted on by a web team — they were designed by someone who ships satellite and RF payloads for a living.
The security gate you'll see below is CMMI Level 4 discipline, scaled to a startup. The founder implemented the most critical components of a CMMI L4 hybrid-agile process on DoD programs using AI agentic workflows — the same decision-record → independent review → field-verification loop now runs VAULTEX.
Someone who has been through accreditation knows what auditors ask for first. That's why the ICD, the ADRs, and the triaged gap list already exist — they're the artifacts an engagement starts with, not what gets written after.
Desktop, mobile, and server all speak the same wire protocol — because they all run the same Rust crypto core.
Interfaces between all 5 platforms × 5 operating modes are formally specified in the Interface Control Document — see SEC_13.
| Primitive | Purpose | Security benefit |
|---|---|---|
| X3DH | Signal-spec session establishment | Forward secrecy from first message |
| Double Ratchet | Per-message key derivation | Post-compromise security |
| XChaCha20-Poly1305 | Authenticated encryption (AEAD) | 192-bit nonces, no reuse risk |
| Ed25519 | Identity + request signing | Sovereign cryptographic identity |
| Sealed Sender | Anonymous authenticated encryption | Server can't identify sender |
| Argon2id | Password / PIN hashing + KDF | Memory-hard brute-force protection |
| HKDF-SHA256 | Key derivation from DH outputs | Secure key expansion |
A timed walkthrough of the VAULTEX chat screen, with the relay's own view underneath. The point isn't the bubbles — it's the log.
Simulation for this briefing — same screens, same envelope shape, scripted timing. The real app, real relay, real crypto: vaultexchat.org →
Every release below is a tagged build with installers on GitHub Releases, auto-served from the website. The cadence is the signal.
Releases are cut from GitFlow main, built locally per platform, published to GitLab and the public GitHub mirror in the same step.
Every feature below operates on end-to-end encrypted data. The server never sees plaintext.
WebRTC signaling over an encrypted WebSocket. SDP and ICE candidates encrypted end-to-end — the server relays without inspecting. Offer → Ring → Connect state machine, missed-call history, mute/speaker controls.
HD 1080p calling with screen share, built on the same signaling path as voice. Windows and Android hardware-verified end to end; Linux media blocked on WebKitGTK and tracked as a known gap.
The same audited crypto runs everywhere — proven by cross-platform field testing on real devices, not just unit tests.
Built for disaster response, press freedom, and conflict zones — contexts with no trusted network at all.
Default online transport. WebSocket + HTTP fallback.
SOCKS5 proxy through Tor. IP-level anonymity.
VXIDv2-authenticated dual-port handshake. Windows↔Linux↔Android field-verified over a real router, zero server involvement.
Direct device pairing — byte-exact verified iPhone↔Android.
Peer-to-peer, no router needed. Android today; iOS equivalent tracked.
Store-and-forward, multi-hop. Signed SOS/discovery beacon for disaster relief — shipped in v0.16.0.
Every platform, every mode, formally specified — not assumed. 14 tracked interop gaps, triaged by severity, visible to us before they're ever visible to an auditor.
| Mode \ Platform | Windows | macOS | Linux | Android | iOS |
|---|---|---|---|---|---|
| Server Relay | ✓ | ✓ | ✓ | ✓ | ✓ |
| LocalNet (LAN) | ✓ | ✓ | ✓ | ✓ | ✓ |
| WAN P2P | ✓ | ✓ | ✓ | ○ | ○ |
| WiFi Direct | ○ | ○ | ○ | ✓ | ○ |
| Bluetooth LE | ○ | ○ | ○ | ✓ | ✓ |
Hundreds of automated tests across Rust, Kotlin, Swift & TypeScript — incl. Signal-spec vectors + adversarial security tests.
One founder, paired with AI coding agents, shipping safety-critical cryptography across five platforms — running a CMMI Level 4 hybrid-agile discipline at startup scale. What follows is as much a case study in AI-augmented software engineering as it is a product demo.
A written decision record for every crypto or transport change
Implemented against the ADR
Independent reviewer re-derives the crypto/auth logic, writes adversarial tests
Unit + integration + Signal-spec vectors
Real devices, real network, end-to-end
Required before merge — no exceptions for crypto/transport
BLE and WiFi-Direct have no emulator equivalent — these flows only exist by testing real phones. The same discipline that catches interop bugs unit tests can't.
This is what lets AI-agent-driven development be rigorous instead of reckless — an AI agent and a human reviewer reason from the same source of truth.
Direct, privacy-conscious users. Open-core — the crypto crate is public and auditable. Premium features & support as the paid tier.
License the zero-knowledge messaging backend + mobile SDKs to companies that need secure comms as a feature of their product.
Market benchmark: a public-safety dispatch/radio software vendor reportedly pays ~$25K/yr for backend messaging at ~50K users — roughly $0.50/user/year wholesale.
Regulated industries — healthcare, legal, government. Zero-knowledge comms under their own infrastructure control. License + support contract.
A developer platform: other teams put VAULTEX end-to-end encryption inside their own product — their UI, their web portal, or their servers and AI agents sending encrypted alerts to people. Design complete; not yet built.
Pricing is the open question: commodity chat APIs run about $0.60/user/yr, while E2EE vendors charge $3–$35/user/month. A design-partner pilot decides which side VAULTEX lands on.
Companies already pay chat vendors to build messaging into portals and apps, and those vendors can read every message. This puts the same convenience on top of end-to-end encryption, including for automated senders such as dispatch systems and LLM chatbots. Design: ADR-0012.
Their UI, our crypto. Native Android and iOS, desktop, and a reduced-capability web build. Target radios run Android, so the Android SDK comes first.
Servers and AI agents become real encryption endpoints, with keys held in the customer's own KMS or HSM. Alerts carry acknowledgement, expiry and de-duplication.
Control plane only: tenants, enrollment, signed organisation badges, webhooks. It never sees plaintext or private keys.
| Pricing anchor | Per user / year | Users for $1M ARR |
|---|---|---|
| Commodity chat API (e.g. Twilio Conversations) | ~$0.60 | ~1.7 million |
| E2EE embed, entry tier (e.g. Element) | ~$36 | ~28,000 |
| Enterprise E2EE (e.g. AWS Wickr) | ~$60–$420 | ~2,400–17,000 |
Competitor prices are public list prices gathered October 2026 and should be rechecked before quoting. These figures are arithmetic, not a forecast.
Head-to-head
| Capability | VAULTEX | Signal | Telegram | Typical enterprise vendor | |
|---|---|---|---|---|---|
| No phone number required | ✓ | ○ | ○ | ○ | ✓ |
| Zero-knowledge server (no contact graph, no metadata) | ✓ | partial | ○ | ○ | ○ |
| Sealed sender | ✓ | ✓ | ○ | ○ | ○ |
| End-to-end encrypted by default | ✓ | ✓ | ✓ | ○ | partial |
| Open-source cryptography | ✓ | ✓ | ○ | partial | ○ |
| Works with no internet (LAN / BLE / mesh) | ✓ | ○ | ○ | ○ | ○ |
| Licensable, embeddable backend | ✓ | ○ | ○ | ○ | ✓ |
Competitor cells reflect publicly documented behavior at the time of this briefing. "partial" = available but not default, or not complete. Signal gets credit where it's due; the gap is off-grid and licensability, not cryptography.
Use of funds
A focused crypto/protocol engagement on the shared Rust core from a specialized firm (Trail of Bits / NCC Group / Cure53-class).
Close the known gaps, freeze the release candidate, and turn the internal design docs into an audit-ready whitepaper.
Remediate what the audit surfaces, cut the v1.0 public release, and land a first design-partner pilot in public safety or humanitarian relief.