Own the means of intelligence.

Reclaim control.

Sovereign AI systems that turn impossible
into the inevitable.

Concept footage. Synthetic illustration.

Thesis

Intelligence is becoming infrastructure. Whoever controls it shapes every decision that follows. We build models, platforms and workspaces designed to keep that control with the people who depend on them.

One stack.
Every layer yours.

Focused models for defined decisions. A platform to reach them. A workspace to direct the work. Each built to be understood before it is trusted.

  1. Praxis Decision NanoDecision modelBeta
  2. consoleAPI platformHosted
  3. harnessAgent workspaceLive
  4. noecodeCoding agentIn development

From signal
to decision.

A concept visualization. The scenario is synthetic and the values are illustrative.

  1. 01

    Detect

    Imagery, reports and signals surface what changed.

  2. 02

    Rank

    A model weighs the evidence and ranks the candidates.

    TRK-01
    0.91
    TRK-02
    0.64
    TRK-03
    0.38
  3. 03

    Approve

    A person decides. Approval always precedes tasking.

  4. 04

    Task

    The approved action is assigned to the right system.

  5. 05

    Confirm

    The outcome returns and updates every connected entity.

Models.
Built for decisions.

Praxis is our family of decision models. Decision Nano is the first release, available in beta for experimentation and evaluation.

Praxis familyBeta, first release

Praxis
Decision Nano

Context in.
A decision out.

Select an option. Estimate a probability. Score a defined outcome. Performance varies by task, so the results are published and you can check them against your own data.

Decision formatsRequest example

Select an option

{
  "type": "choice",
  "state": {
    "message": "I paid twice for the same subscription."
  },
  "question": "Which team should handle this ticket?",
  "options": {
    "billing": "Payments, charges and invoices",
    "technical": "Software errors and connectivity"
  }
}
Output fieldschoice · probabilities · confidence

Example payloads only. No live inference.

Model ID
praxis-decision-nano
Formats
Choice, Noul, Score
Input limit
2,048 tokens
Input pricing
$1 / M tokens
Top-answer accuracy, all questions
59.70%Confirmation evaluation on 2 Oct 2026, 4,682 questions

Products.
Tools that do the work.

Focused tools for putting intelligence to work: reach hosted models, coordinate agents, and build close to the system.

Swipe to see all three

API platformHosted platform

console

Manage API keys, explore available models and track usage. Inference requires prepaid credit; credit purchases are not available yet.

Open console
Agent workspaceLive workspace

harness

Models, tools and context on one operating surface. Bring your provider; your workspace data stays in this browser.

Explore harness
Coding agentIn development

noecode

A terminal-native coding agent, designed to work inside your own environment. Navigate, plan and build close to the system.

Follow the engineering

From intent.
To execution.

A model alone is not a working system. Useful work needs context, a deliberate plan, and a result someone can inspect.

Illustrative workflow

“Understand this repository.
Plan the change. Verify the result.”

From a repository to a reviewable change
ContextExample artifact

Before action, understanding.

Read the structure, trace the dependencies, and identify the constraints. Give the work a foundation beyond a single prompt.

Context brief
  • Map entry points and dependencies
  • Identify conventions and boundaries
  • Surface missing information
Human judgment at every transition.A system design study, not a live product session.

Your intelligence.
Your terms.

The ability to use AI should come with the ability to choose how it operates. We are building toward practical AI sovereignty: one system at a time.

Read the mission
  1. Data

    Control your data.

    Decide where knowledge lives, who can access it, and how it is used.

  2. Models

    Choose your models.

    Use the models that fit the work, with the freedom to switch as needs change.

  3. Systems

    Own your environment.

    Choose where intelligence runs, with a path toward infrastructure you control.

  4. Decisions

    Keep people in command.

    Inspectable actions, explicit permissions, and human judgment where it matters.

The whole system.
Not just the interface.

Compose

Systems over features.

Build durable foundations that can be understood, connected and extended.

Ship

Research into reality.

Turn promising ideas into useful tools, tested against the work they need to do.

Reduce

Less, but deliberate.

Reduce friction while keeping the controls that matter visible and within reach.

Engineering, in the open.github.com/noeraven

Build something
difficult.

Building toward greater control over your AI systems? Bring us the problem worth solving.

Start a conversation