OpenCognit
Orchestrator

One sentence,
a whole team at work.

Describe what should happen. The orchestrator starts suitable agents, gives each its own files, checks their work with your tests and reports back — on your own subscriptions, on your own computer.

Claude Code · Kimi Code · HermesLinux & WindowsNo telemetry
See it run

From one sentence
to a running team.

What you get

A conductor that keeps
the team in order.

I

Free agents first

It uses agents that are idle before starting new ones — and you confirm large start requests beforehand.

II

Every task has a number

Each task moves through fixed stages: queued, delivered, working, checking, reported. So the orchestrator always knows where each agent stands.

III

Agents report back

Each agent ends with a report — done, blocked or question — plus a summary and the changed files.

IV

Own files for each agent

When several agents work in one project, each gets its own files or folders. Changes outside them are blocked, so twelve agents don’t get in each other’s way.

V

Tests before “done”

When an agent reports done and the project has tests, OpenCognit runs them. If they fail, the output goes back to the same agent — and a test guard notices weakened tests.

VI

Keeps going at the limit

If an agent hits its usage limit, the task moves to another account or another signed-in CLI, with the same number and state.

How it works

Three steps to a team.

I

Open the orchestrator

In the side panel. It runs on Claude Code, Kimi Code or Hermes.

II

Say what you want

One sentence is enough — it plans, starts agents and hands out the work.

III

Decide at the end

A card per agent shows the changed files with +/− lines: accept, refine or discard.

Questions about the orchestrator

Which agents can it start?

Claude Code, Codex, Gemini, Kimi Code and Hermes — whatever you have installed. It can conduct with Claude Code, Kimi Code or Hermes.

Does it change my code by itself?

The orchestrator only reads. The agents you see make the changes, and risky commands still wait for your approval.

What happens after a restart?

Open tasks aren’t lost: waiting ones get delivered, and every agent continues its own conversation.

How many agents can run?

You set the limit under Settings → System. Large start requests need your confirmation.

What does it cost?

Free with up to 4 agents and 2 projects. Pro is a one-time purchase — no subscription on top of your plan.