● [ SOVEREIGN AI INTELLIGENCE • ZERO CLOUD ]

WHERE PRIVATE MINDS

OUTPACE THE CLOUD

POWERED BY IRIS

( A )

[ 43.9 T/S • 0MS CLOUD PING ]

Iris unifies 100% offline quantized foundation models (Llama 3.2, Qwen 2.5, DeepSeek-R1), real-time speech dictation, and drop-in OpenAI REST endpoints into one sovereign workstation. Absolute privacy. Zero monthly subscription costs.

Launch Workstation

/ ABOUT IRIS SOVEREIGNTY

Engineered for individuals and organizations who refuse to lease their intelligence from closed corporate silos.

Every token generated, audio waveform transcribed, and memory vector indexed executes directly on your bare-metal processor. Iris gives you sovereign, uncensored, zero-latency artificial intelligence with complete data isolation.

[ 01 / VOICE ]

Talk Now

Speak Live Now

Two-way hands-free conversational voice powered by local Whisper STT, an interactive 3D WebGL reactive orb, and low-latency Kokoro/eSpeak neural speech synthesis.

[ 02 / ENGINE ]

Explore Engine

Local Sovereign Engine

Native GGUF acceleration with Ollama backend. Run Llama 3.2, Qwen 2.5, and DeepSeek-R1 at 43.9+ tokens/second right on your CPU or GPU without cloud fees.

[ 03 / API ]

Build API

Build Something New

Drop-in OpenAI REST API on http://127.0.0.1:1337/v1. Seamlessly connects to Cursor, Claude Code, Cline, and custom autonomous multi-agent pipelines.

IRIS®

[ ESTABLISHED 2026 • IRIS AI ]

/ MISSION INTELLIGENCE:

UNRESTRICTED SPEED FOR ANY INNOVATION

OFFLINE GGUF WEIGHT ACCELERATION WITH ZERO CLOUD DEPENDENCY.

Deploy state-of-the-art weights on your host without risk of data harvesting, API rate limits, or unexpected billing spikes. Iris runs fully isolated inside your infrastructure, protected by NeMo guardrails and local vector memory.

Learn about Iris AI

LLAMA 3.2 QWEN 2.5 DEEPSEEK-R1 MISTRAL OLLAMA COOLIFY HETZNER CPX CURSOR CLAUDE CODE IRIS AI LLAMA 3.2 QWEN 2.5 DEEPSEEK-R1 MISTRAL OLLAMA COOLIFY HETZNER CPX CURSOR CLAUDE CODE IRIS AI

[ RESPONSIBLE SOVEREIGN AI & LEGAL GOVERNANCE ]

The Foundation of Trust: Do's, Don'ts & Absolute Sovereignty

Iris is designed to empower human agency through decentralized computing. In alignment with frontier industry safety benchmarks and responsible open-source AI principles, our operational charter establishes strict boundaries for deployment and data protection.

Permitted & Encouraged Use

✓ Run private enterprise research, proprietary document analysis, and confidential strategic planning with zero external exposure.
✓ Build offline developer tools, automated code generation, and multi-agent systems via the localhost OpenAI server.
✓ Experiment freely with quantized foundation models (Llama, Qwen, DeepSeek) for educational, scientific, and commercial productivity.
✓ Implement custom safety rules and deterministic guardrails using Iris NeMo policy configuration.

Prohibited Actions & Misuse

✗ Do not deploy Iris for malicious cyber intrusions, automated credential harvesting, or payload synthesis.
✗ Do not generate deceptive non-consensual synthetic media, impersonation content, or mass disinformation campaigns.
✗ Do not represent unverified synthetic responses as certified medical diagnoses, legal counsel, or financial fiduciary recommendations.
✗ Do not deploy for real-time life-safety critical control systems without human-in-the-loop validation.

PRIV

Zero Telemetry & Data Isolation

● No External Network Calls: In local inference mode, prompts, voice recordings, and weights never traverse the internet.
● Zero Diagnostic Telemetry: Iris collects zero behavioral telemetry, analytics, or user keystroke logging.
● Self-Contained Storage: User memory threads, scheduled tasks, and embeddings reside exclusively in local SQLite database files (./data/iris.db).
● Complete Deletion: You retain 100% ownership. Deleting your workspace instantly wipes all local vectors and histories.

LEGAL

Terms of Service & Model Licensing

● Open Foundation Weights: Subject to Meta Llama 3.2 Community License, Alibaba Qwen License, and Apache 2.0 open-source terms.
● Intellectual Property: All generated code, artifacts, and creative output produced by the local workstation belong entirely to the user.
● Warranty Disclaimer: Software is provided "as is" without warranty of continuous uptime or hallucination-free output.
● Support & Customization: Professional customization and deployment backed by Iris AI (irisai.live).