“An AI that cannot show its work is not intelligence — it is liability.”
The engine every KI module runs on. Not a chatbot. Not a cloud API. The sovereign inference runtime — constitutional pre-flight, sovereignty enforcement, on-device or KI-hosted. Browse modules in RigidMarket. Every one of them runs here.
Every module in the RigidAI platform carries a sovereignty indicator. It tells you exactly where inference runs today — and where it is going. No other AI platform shows you this. We do because the roadmap is the product.
Runs fully local on your RigidNode hardware. No inference crosses your network. No data leaves. The Sovereignty Indicator stays green.
Local storage. Cloud inference. Every external call is logged in an immutable audit trail in RigidVault. You see exactly what left and when.
Cloud inference today. Migration path defined. Every Assisted module has a version on the roadmap scheduled to flip Sovereign as local models mature.
Customers on Sovereign Only mode see only 🟢 modules activated. As modules flip from Assisted to Sovereign, they unlock automatically — no action required. See RigidTrust framework →
Before any RigidAI module takes a consequential action — writing data, transmitting data, controlling hardware, generating output a human will act on — three checks execute in sequence. Not as guidelines. In code. In milliseconds.
The action is checked for physical, financial, privacy, and health harm. If it fails, the action is blocked before it executes, logged permanently, and the user is shown a plain-language explanation of what was stopped and why. The block cannot be overridden by any user, admin, or configuration.
The action is checked against the user’s explicit request scope. If the module is about to process more data than was asked, take a more consequential action than was requested, or act autonomously beyond the stated task — it stops and asks for confirmation before proceeding.
The sovereignty log, audit trail, and provenance chain are protected from modification. No module, user, or administrative action can alter a logged event after it is written. The audit trail is the product.
Most AI platforms train on your data by default and require you to opt out. RigidAI inverts this. The default is zero contribution. You activate it deliberately.
No RigidVault asset ever contributes to model training without an explicit opt-in authorization from the vault owner. The default for every new vault is non-contributing. Activation is a deliberate action, not a buried setting.
When a vault owner opts in and their data is used in a training batch, RigidRoyalty calculates a proportional payment based on data type, domain relevance, and contribution weight. The model improves. The contributor earns.
Every training cycle undergoes a third-party ethics audit. Published results. No exceptions for proprietary data. The audit covers consent verification, contribution accounting, and output attribution.
Every RigidAI module belongs to one of 14 families — each family a category of intelligence, each module an app you can install. Browse by family to find what fits your operation. Every module runs on the same sovereign architecture, reads from the same vault, and answers to the same constitutional standard.
Browse all 140 modules in the store →Every RigidAI module displays a Sovereignty Indicator after every operation. The indicator reads from the system log — not from marketing copy. It reflects what actually happened.
SOVEREIGN: All processing on KI hardware. No data left the network. ASSISTED: Storage local, inference used an external model. Audit log shows external API call. CLOUD: Significant processing occurred externally. Full external transmission logged.
Customers in Sovereign Only mode see modules grey out when local inference is unavailable — that greyed module is the Sovereign Node product selling itself.
Highest quality. Claude API inference. Data transmitted to process this request. Audit log shows external call. Right for: development, consumer heritage, small business document work. Fine for most use cases.
Data never leaves the network. Inference runs on KI-class edge compute locally. Quality lower than cloud on complex tasks, excellent for classification, extraction, and summarization. Free, unlimited, sovereign.
Fine-tuned model trained on KI domain data. Runs on KI hardware. Domain-optimized — outperforms general models on CNC diagnostics, heritage documentation, and clinical documentation. The 12-18 month build target.
Switching tiers is one configuration change. No rebuild. No data migration. The module stack does not change — only where inference runs.
The generative engineering division of RigidAI. LLM-driven CAD geometry, toolpath generation, and design variant exploration. RigidForge takes a brief and returns manufacturable geometry — no CAD operator required for the generation step.
14 health modules across three areas: clinical documentation for behavioral health networks, sovereign personal health records, and shop floor worker wellness. The same platform, applied to healthcare.
Energy intelligence — facility-scale power monitoring, peak demand reduction, and renewable integration. Scales RigidSense power sensing from machine-level to the entire facility.
140 modules across 14 families. Filter by pillar, tier, and sovereignty. Install what you need to your RigidNode.
Open RigidMarket →No AI platform is born fully sovereign. Every module starts somewhere on the inference spectrum. RigidAI is honest about where each module runs today — and committed to where it is going. As local models improve and the RigidAI kernel is built, modules migrate from ASSISTED to SOVEREIGN one by one. Each flip is a product milestone customers on Sovereign Only see in real time.
Claude API for complex reasoning modules. Local retrieval and storage. Data leaves for inference only. Full audit log of every external call.
Open-source models running locally on KI-Forge or Jetson. Classification, summarization, extraction modules flip to sovereign. Data never leaves. Free, unlimited, local.
Fine-tuned model trained on KI domain data — CNC diagnostics, heritage documentation, defense specs, engineering standards. Outperforms general models on your specific tasks. Fully on-device.
Every customer on Sovereign Only receives a notification: “[Module name] is now fully sovereign on your vault.” No announcement. No fanfare. The indicator goes green and the capability is there. That notification is a product launch.
The intelligence that grows with every node, every machine, and every consented data point in the ecosystem.
Every pillar in the KI ecosystem is interconnected. No product works alone.
View Full Ecosystem Map →Kavanagh Industries · Always on