ProductsStart from your constraint

What can't
your AI do?

Can't leave your perimeter. Can't work in your language. Can't explain its decisions. Eight systems that solve for those constraints, handed over whole.

Sovereign Infrastructure

04 systems

01 / 08

Cordon

Private Inference Engine

Open Cordon →Source

Exfiltration is architecturally impossible.

The sovereign foundation every other Regnant system runs on. No data, query, weight, or result ever crosses the institution's perimeter. Enforced at hardware, firmware, OS, and application level, from a hardware root of trust to a client-verifiable, Merkle-chained audit log.

TPM 2.0 attestation and Secure Boot: containment enforced in hardware, not policy
AMD SEV-SNP, Intel SGX v2, or ARM TrustZone: inference runs in memory no operator can inspect
AES-256-GCM weights under a key hierarchy the institution alone holds. The vendor cannot decrypt
Ed25519-signed Merkle audit chain, verifiable by the client with no internet connection

“Our AI cannot touch the public cloud. What do we run?”

02 / 08

Knott

Workflow Orchestration

Open Knott →Source

Automation you can actually be accountable for.

Self-hosted, AI-powered workflow automation with human-in-the-loop decisions, a visual designer, real-time execution monitoring, and a full audit trail. Zero cloud lock-in, one exposed port, every decision recorded with its justification.

100% offline capable with local Ollama models
Automatic escalation to human reviewers when AI confidence is low
Only a single port exposed to the outside world

“Every automated step has to survive an audit. Who approved this one?”

03 / 08

Matta

Enterprise Semantic Platform

Open Matta →

A semantic platform, not a file.

One source of truth for enterprise master data. Matta unifies the customers, suppliers, employees, and contracts scattered across CRM, ERP, and HR systems into a single governed layer. Duplicates resolved, quality enforced, every fact traceable to the system it came from. Ask what your business knows, and get an answer with receipts.

One canonical record per customer, supplier, and employee: duplicates resolved across systems
Every fact carries its source, timestamp, and confidence. Audit-ready by default
Point-in-time answers: the business exactly as it stood on any past date
AI answers that cite their evidence, grounded in your data, never invented

“Our systems can't agree on what the data means.”

04 / 08

SeeP

Auditable Operations Agent

Open SeeP →Source

The agent cannot execute a change.

An always-on operations agent for your fleet, structurally incapable of acting without a signed approval. It investigates with read-only tools, and when it wants to alter something it produces a plan that goes through policy, then a human, then an executor that verifies the authorization independently before running a single step. The agent's tool registry contains no mutating tools, so refusal is the shape of the code rather than an instruction in a prompt.

Approvals are ed25519-signed, single-use, and expire. A captured bundle replays into a refusal
The executing node verifies the authorization itself against a key it pinned at enrollment
Policy as code: two-person rules, change freezes, blast-radius tiers. Rules only ever tighten
Hash-chained audit log: editing one entry breaks every link after it
sovereign = true keeps every token on your own hardware, with no remote failover

“Who is on call at 3am, and what stops it breaking production?”

Decision Intelligence

02 systems

05 / 08

Wallgarden

Autonomous Business Operating System

Open Wallgarden →

The decision arrives before the problem does.

An autonomous executive layer for the whole business. Wallgarden connects to the tools you already run (finance, CRM, operations, communications), watches the numbers continuously, and brings leadership decisions that arrive already researched, costed, and ready to approve. It executes what you sign off, learns from every outcome, and gets measurably better each quarter.

A digital executive team on watch around the clock. No prompt required
Every recommendation arrives with reasoning, financial impact, and confidence attached
One click to approve; routine calls execute automatically within limits you set
A value ledger tracks the money each decision saved or generated

“Information is everywhere. Why are decisions still late?”

06 / 08

IIN

Industrial Intelligence Platform

Open IIN →

Hear the failure before the floor does.

Predictive operations for plants, mines, and utilities. IIN listens to the equipment you already run (sensors, PLCs, SCADA), learns what normal sounds like for every asset, and calls the failure days before it happens, with the work order already drafted. Unplanned downtime becomes a schedule item instead of an emergency.

Connects to the plant as-is: OPC-UA, Modbus, MQTT, and existing SCADA systems
Predictions ranked by production value at risk, not just probability
From alert to work order in one motion. Maintenance moves before the line stops

“Every machine is talking. Who hears the failure coming?”

Engineering

01 system

07 / 08

Bubbly

Self-Hosted Coding Agent

Open Bubbly →Source

Agentic engineering, under containment.

Engineering rigour an institution can actually approve. Bubbly holds a plan, edits across files, runs the suite and reviews its own diff before it asks for anything, on models you host: Ollama on the engineer's machine, Cordon where the work has to be attested. The source never leaves the perimeter and every destructive action stops at a human gate.

Plan, build, test and review in one loop, with the diff reviewed before it is offered
Spec mode: requirements, design, and tasks agreed with you before a line is written
The Bubble Room: mission control over every agent run, with undo on everything
Runs on Ollama locally or on Cordon for attested inference. No external API calls

“Our engineers want AI coding agents. The code can't leave.”

Education: the Milkshake ecosystem

01 system

08 / 08

Milkshake

Sovereign Education Ecosystem

Open Milkshake →Kairos Weave

The whole classroom, in your language, offline.

One education ecosystem with two sides. Kairos carries the teacher's day: lesson plans, NECTA-style exams, marking, and report comments in seconds, in English or Kiswahili. Weave carries the student's: a bilingual study and research workspace where Tanzanian students analyze data, build software, search curated local sources, and render interactive artifacts, all offline. Open-licensed for schools, NGOs, and government education projects.

Kairos, the teacher's half: TIE-format plans, NECTA-style papers, marking and reports
Weave, the student's half: retrieval, real Python on real data, software builds, offline artifacts
Runs on 4GB of RAM and needs the internet once, for setup. No cloud, no telemetry
Generates directly in Kiswahili, not through a translation layer

“What does sovereign AI mean for a school with no internet?”

KW5-Lite, the 109.5M-parameter Kiswahili foundation model, is a research programme and has a page of its own under Research.