OMNIS KEY

Open runtime for local AI agents.

What is OMNIS?

OMNIS is a lightweight, extensible runtime for building and running AI agents locally.

• CLI-first — start, stop, query from the terminal
• Plugin-based architecture — extend with your own tools and integrations
• Multi-agent orchestration — spawn and coordinate multiple agents
• Local-first execution — runs on your machine, your infrastructure

Why OMNIS?

Most agent frameworks are either too abstract, too cloud-dependent, or too rigid.

OMNIS is designed to be:

• Simple — up in 30 seconds
• Hackable — clean TypeScript, easy to extend
• Composable — bring your own LLM, your own tools

Getting Started

npm install -g omnis-key
omnis-key start --dev
omnis-key status

Then open: http://localhost:18800

System requirements: Node.js 18+

Example

# Start the runtime
omnis-key start

# Run a query
omnis-key query "What's the current market signal for Permian Basin?"

# Check status
omnis-key status

# Stop
omnis-key stop

COSMIC Integration (Optional)

OMNIS connects to the COSMIC intelligence layer for advanced reasoning:

• Causal inference (NOVA)
• Probabilistic decisioning (ECLIPSE)
• Adversarial stress-testing (PULSAR)
• Trust + audit systems (AURORA, HEIMDALL)

Available via hosted API at api.omniskey.ai.

# Point to your COSMIC backend
export OMNIS_API_URL=https://api.omniskey.ai
export OMNIS_API_KEY=your_key

omnis-key start

AURORA Hard Gate: Analytical queries with confidence below 0.80 are blocked — not passed through. OMNIS doesn't guess. It decides, or it tells you it can't.

Architecture

OMNIS SHELL (this repo — MIT)
       ↓
   API call
       ↓
COSMIC CORE (hosted intelligence — proprietary)

OMNIS is the runtime. COSMIC is the intelligence.

Philosophy

The world has enough AI tools that tell you things.

OMNIS is built for systems that need to decide things — with a provable reasoning chain, adversarial validation, and a hard gate on low-confidence outputs.

The runtime is open. The intelligence is earned.

License

MIT — JourdanLabs