Skip to content

Platform

Your agents. Your fleet. Independent checks.

The OmniConflux Confluence mark: seven glass strands, each carrying its own light, braided into one ring.

OmniConflux is a desktop coding app for Omarchy, macOS on Apple silicon and Windows, with Mission Control, rival review, Fleet, the Arena (on Omarchy) and the Council. Version 0.3.0 is in closed beta. Local context and voice remain part of the design described below.

Your first lane

Choose the builder, the rival verifier, the acceptance command and what the lane may do before launch.

Launch preview with acceptance settings, four automatic-step controls, and a Launch lane button.
Launch preview chooses which steps run automatically before the lane starts.
Review screen with a Pass rival verdict, test evidence, and a Merge button.
Review records a rival Pass and the test evidence behind it.

Every agent checked by a rival vendor

OmniConflux gives every task a verifier from a different vendor. The agent and the verifier bring separate perspectives to the work. Verification is a check on the result, not a promise that every result is correct.

See how the rival check works
Six machines, phone to server rack, all connected to one glowing hub

Your machines are the cloud

The design pools your own machines into a fleet. Your computing resources are designed to form the foundation of the agent workspace.

  • Private by default

    Local models and on-device context keep the design centered on your machines. These are the starting principles for privacy.

    Read the privacy principles
    A closed glass vault with a round door wheel, a private local model glowing green-white inside
  • Talk to your fleet

    OmniConflux is designed voice-first. The design lets you work with your agents through the fleet you own.

    Explore voice
    A studio microphone with a glowing waveform ribbon flowing out of it
  • Earning

    Aura by OmniConflux is the platform's ad network, running today for approved closed beta accounts. Half of what each ad earns goes to your Aura wallet.

    Learn about earning
    An ad card with a clearly legible Sponsored label and a violet coin leaning on it

Any vendor

Open by design

OmniConflux is designed to work with any vendor and never pigeonhole you. Here is exactly what that means, so nothing is left to the imagination.

What we say, and what it means precisely

OAuth 2.1 and OpenID Connect sign-in, no long-lived keys in your agents

The design uses the Authorization Code flow with PKCE, exact redirect matching, header-bound tokens and refresh token rotation for accounts and connectors. Connectors are designed to publish a client metadata document, so you can see exactly who is asking for access.

Bring your own models: subscription, API key or local

Subscription: OmniConflux is designed to run the vendor's own installed tool, such as Claude Code, Codex CLI or Gemini CLI, on your machine under your login.

API key: your key, direct to the vendor.

Local: Ollama or llama.cpp on your own hardware.

Any harness

Designed for process-level control of any command-line agent, with deeper control, including state, handoff and approvals, for agents that speak MCP or ACP. Agents are designed to share work through git worktrees and files, with no claim of perfect memory across vendors.

Cross-vendor verification

A verifier from a different vendor family checks each diff before it counts as done.

Private by default

When you choose a local model, the design keeps the work on your machine.

When you choose a cloud model, that vendor receives what the agent sends it.

Self-serve is the default design. A managed fleet is planned and not live; to discuss one, talk to us. It will run on your API keys or enterprise tokens, never on personal subscription seats. Read why on the digital sovereignty page, and see the full list on the integrations page.

Start with the concepts

Read how the pieces fit together, or pre-register for the closed beta. Nothing can be bought on the site today.

Questions

Can I use my existing AI subscription?

On your own machines, that is the design: OmniConflux runs the vendor's own tool under your login, as the vendor allows. We do not run many cloud agents on one personal seat, pool personal accounts, or pass your seat login through a hosted backend.

Does every agent work with zero setup?

No, and we will not claim it. Any command-line agent is designed to run at the process level. Agents that speak MCP or ACP are designed to get deeper control. Some agents need a short setup step.

What is the difference between self-serve and managed?

Self-serve is the default design: your machines, your keys, your subscriptions on your own machines, and local models. A managed fleet is planned and not live; to discuss one, talk to us. It will run on your API keys or enterprise tokens, never on personal subscription seats.

All product names are trademarks of their owners. Naming a product describes designed compatibility, not endorsement.