Ordinis — Home

Ordinis AI

Assistance that shows its work.

Ordinis AI is the intelligent interface to the Ordinis ecosystem. You describe what you need; several AI models plan it, do it and check each other’s work; and you approve anything that matters before it happens.

The Ordinis AI console showing one task worked on by three AIs: ChatGPT plans, Claude checks the plan and builds, Codex reviews and passes it, and Claude verifies the result.
One task, three AIs, one answer.

Who it’s for

For work that deserves a second opinion

People who build with AI

Plans, code and documents produced by one model and checked by another, in one place.

People who want the last word

Anything that ships, publishes or deploys waits for a person to approve it.

Long-running projects

Each project keeps its purpose, its decisions and its results, so work picks up where it stopped.

How it works

Several models, one answer, your approval

Instead of trusting one model, Ordinis AI lets models with different strengths work as a team, and keeps you in the loop at the moments that count.

You askIn words, typed or spoken
A model plansThe task is broken into steps
A model buildsThe work is done in a scoped workspace
Another model reviewsA second opinion checks the result
You approveNothing important ships without you

Conversational

Ask in plain language. Ordinis AI turns your request into a plan you can read before anything runs.

Reviewed by default

One model builds, a different model reviews. You get one answer, with the review attached.

Approval for anything that ships

Publishing, deploying or changing something important always waits for your approval.

Project memory

Notes and results are kept per project, so the next task starts from what was already decided.

Model choice

Models are interchangeable providers, not a single vendor. Today it works with ChatGPT and Codex from OpenAI and Claude from Anthropic.

Guardrails

Work runs with a chosen access level (read only, one project, or custom). Guardrails reduce risk; they are not a sandbox, and we say so.

The console

Every step visible. Every decision yours.

A stepper follows each task from plan to memory: Plan, Check plan, Work, Review, Verify, Remember. The full transcript and timeline are one click away.

An Ordinis AI task with its stepper Plan, Check plan, Work, Review, Verify and Remember, and the verified result.
A finished task, verified and remembered.
The Waiting for you page with decisions and approvals, each with Approve, Request changes and Decline.
Approvals and decisions wait in one place.
The Projects page with a card per project showing its status, purpose and next action.
Every project, its purpose and its next step.

Principle

AI proposes. People decide.

Model output is a proposal; approvals come from people. Nothing reaches production without your approval, and the dashboards show real usage only where a provider reports it: they never guess.

Privacy, permissions and security

Built to ask, not to assume

Local versus cloud

Requests are processed by the cloud AI providers you connect. We always say which parts of a feature run where.

Permissions

Sensitive actions (sending, deleting, publishing, changing system settings) require explicit approval. Guardrails are designed to keep the models away from stored credentials.

Audit

Each task keeps its plan, its review, its approval and its full transcript.

No hidden autonomy

Ordinis AI does not act on your accounts or devices on its own initiative. It works only with the access level you choose (read only, one project, or custom).

Under the hood

Technical architecture

Console
Web console with a local service; Windows app and Linux user service
Models
Third-party models via the user’s own subscriptions (OpenAI, Anthropic)
Integration
Model Context Protocol (MCP) so tools and models can call each other
Governance
Approval lane for anything that ships; full transcripts per task
Platforms
Windows and Linux desktops

Questions

Is Ordinis AI its own AI model?
No. Ordinis AI coordinates models from other providers and adds planning, review, memory and approvals around them.
Can it act on my computer by itself?
No. Anything that changes something important waits for your approval.
Is it private?
Requests are processed by the cloud AI providers you connect, and every task keeps its full transcript so you can see what was sent.
Which AI models does it use?
Today ChatGPT and Codex from OpenAI and Claude from Anthropic, through the user’s own subscriptions. Models are interchangeable providers, not a single vendor.
Is it the same as the Ordinis AI icon on the Ordinis OS dock?
No. That icon opens a small chooser that starts the assistant you pick in Ordinis GO. Ordinis AI, on this page, is the console where several models work on a task together. See the chooser.
How does it relate to Odin?
Ordinis AI is what you use. Odin is the governance platform designed to authorise and record every step. Ordinis AI follows the same rule with its own approval lane.

Part of the family

How it works with the other Ordinis products

Odin is the governance platform designed to authorise and record every step; Ordinis AI follows the same rule with its own approval lane.

Runs as a user service on Linux desktops such as Ordinis OS, and as a Windows app.

Ordinis Lens applies the same principle in the browser: AI reads the sources, code checks the quotes, you draw the conclusion.

Recent milestones

What’s new

  • Console 2.0 · A stepper for every task: Plan, Check plan, Work, Review, Verify and Remember.
  • Waiting for you · Approvals and decisions in one place, with Approve, Request changes and Decline.
  • Projects · A card per project with its status, purpose, next action and where that information came from.
  • Honest dashboards · Activity per AI, and quota only where the provider reports it.

Support us

Support Ordinis

Ordinis is independent. If an app helps you, you can support its development.

Support us

Payments are handled securely by Stripe.