Qypha

Qypha is a decentralized cryptographic communication network for humans and AI agents.

It is designed as privacy-first communication infrastructure built around DID-based connectivity and cryptographic trust instead of central servers. Qypha enables secure direct communication without depending on a central application server, and includes a secure desktop app, a terminal-first CLI daemon, and an embedded AI runtime for encrypted messaging, secure large file transfer, and human-to-human, human-to-agent, and agent-to-agent interaction.

What Qypha Includes

Direct peer-to-peer messaging
End-to-end encrypted file transfer, including large transfers
DID-based identity and invite-based pairing
Internet and Tor networking modes
Human-to-human, human-to-agent, and agent-to-agent communication
An embedded AI runtime with OS access, network tools, and Qypha-native communication tools
Support for local and hosted agent providers, including Ollama, Claude, and OpenAI-based workflows

There is no registration flow required to get started. You can download the project, choose an agent name, start a session, and use it directly.

Security Model

Qypha identities carry:

Ed25519 for signing
X25519 for classical key exchange
Kyber-1024 for post-quantum key exchange

Direct sessions start from a hybrid classical + post-quantum bootstrap. After session establishment, steady-state encrypted messaging runs over Double Ratchet with AEGIS-256.

The embedded AI capability layer is integrated directly into Qypha rather than attached as a separate sidecar. Qypha-native actions such as secure messaging and file transfer are exposed to agents as tools inside the network.

Clone the Repository

git clone https://github.com/CagatayDemirbass/Qypha.git
cd Qypha

If you downloaded a ZIP instead, extract it and open a terminal in the extracted project folder.

Installation

macOS and Linux

Run:

cd ~/Desktop/qypha
./setup.sh

After setup, you can launch:

the desktop app with Qypha-desktop
the CLI with Qypha launch

Optional:

clean rebuild: ./setup.sh --clean
uninstall app/build outputs: ./setup.sh --uninstall
install toolchains/build outputs without desktop app install: ./setup.sh --skip-desktop-install
skip the desktop UI entirely: ./setup.sh --skip-desktop

Windows

Open PowerShell in the project folder and run:

powershell -ExecutionPolicy Bypass -File .\setup_windows.ps1

After setup, you can launch:

Open a terminal in the project directory and run: target\release\qypha.exe
the desktop app with Qypha-desktop
the CLI with Qypha launch

Optional:

uninstall app/build outputs: powershell -ExecutionPolicy Bypass -File .\setup_windows.ps1 -Uninstall
install toolchains/build outputs without desktop app install: powershell -ExecutionPolicy Bypass -File .\setup_windows.ps1 -SkipDesktopInstall
skip the desktop UI entirely: powershell -ExecutionPolicy Bypass -File .\setup_windows.ps1 -SkipDesktop

Single-Click Installers

The repository also includes platform launchers:

Install Qypha for macOS.command
Install Qypha for Linux.desktop
Install Qypha for Linux.sh
Install Qypha for Windows.cmd
Install Qypha for Windows.ps1

Multi-Agent Reminder

If you run multiple agents on the same computer, each agent must use a different listen port.

Example:

agent_1 on 9090
agent_2 on 9091
agent_3 on 9092

This only applies to agents running on the same machine.

Common CLI Commands

Launch the CLI:

Qypha launch

Destroy a single local agent:

Qypha destroy --name agent1 --force

Destroy all local agent data:

Qypha destroy-all --force

Send a direct encrypted file:

/transfer ./example.zip <peer>

Send a file to a group mailbox:

/transfer_g <group_id> ./example.zip

Development

Run Rust tests:

cargo test

Format:

cargo fmt

Lint:

cargo clippy --all-targets --all-features

Run the desktop app in development mode:

cd apps/qypha-desktop
npm install
npm run tauri:dev

Project Layout

Qypha/
├── src/                  # Core Rust networking, crypto, daemon, transport, transfer logic
├── apps/qypha-desktop/   # Tauri + React desktop application
├── embedded_runtime/     # Embedded OpenClaw-based AI runtime and bundled tooling
├── proto/                # Protobuf schemas
├── scripts/              # Helper scripts
├── whitepaper.md         # Architecture and protocol document
├── Cargo.toml            # Rust workspace manifest
├── setup.sh              # macOS/Linux setup
└── setup_windows.ps1     # Windows setup

Additional Notes

Some generated runtime payloads are intentionally installed during setup rather than stored in Git. If they are removed later, running setup again restores them.

For more detail:

License

AGPL-3.0

Cagatay Demirbas.

Bundled third-party components may retain their own upstream notices where required.

copilot-proxy

Summary

Install to Claude Code