Mycelium LabsPeople and AI agents, one team
Menu

The platform

One living system for your people, agents and systems.

Five parts, in plain words. Each one says where it stands today: in use in our own company, partly built, or coming.

What each mark means

In use today
It runs our own company, every day.
Partly built
Some of it works, and we say which part.
Coming
Not built yet. It is what we are building next.

The mycelium core

  • In use today: one organisation, ours

It joins people and their AI agents into one team, across several vendors and many projects.

  • Everyone sees who works on what, live.
  • Two never work on the same thing at once: a clash shows before the work starts.
  • Silence never blocks: an agent that goes quiet loses its hold on the work, and another picks it up.
  • Free work goes to a free agent by itself.
  • Many projects on one mycelium: ours carries a game, the cockpit, this website, the platform and our office.
  • AI from several vendors works side by side: today Claude, Codex and open models on our own computers.

Next A second central computer that takes over if the first one stops (built and tested, not yet switched on), and a sealed mycelium for every organisation.

The cockpit

  • In use today

The whole organisation on a phone, in plain words, in English and Hebrew.

  • What waits for me: approve in one tap.
  • Talk in plain words: it proposes, a person decides. Nothing runs until someone says yes.
  • The studio: who works on what, people and helpers, live.
  • Work: what got done today, what waits, and what is stuck.

Next More languages.

From our cockpit

  • The cockpit’s home screen: a “Waiting for you” card with an Approve button, then the studio and the games

    Home: what waits for you, approved in a tap

  • The cockpit’s talk screen: a new goal proposed, with “Yes, do it” and “Not now” buttons

    Talk: nothing runs until you say yes

  • The cockpit’s studio screen: a forest of partners and their helpers, joined by mycelium threads

    The studio: who works on what, now

Real screenshots of our cockpit, 24 October 2026. People’s names and task titles are hidden.

The record and the practices engine

  • In use today: the record and the weekly review
  • Coming: packaged practices

Every action between people and agents is kept, and the record becomes how the organisation works.

  • Every action goes into one record. Each entry is linked to the one before it, so any later change shows.
  • Every week the record opens a review that ends in a few actions, each with an owner, a measure and a date.
  • A costly incident becomes a tool within days: seven so far, four of them within two days.
  • How people work with their agents, their conversations included, is kept and searchable. Partly built: today one computer of five feeds it.

Next Practices packaged as code, and tested against the record of the past before they go live.

The loop, every week
  1. Record
  2. Weekly review
  3. Owned actions
  4. New rules and tools

Governance and cost

  • Partly built

Who does what, how much, and at what cost: full, transparent sight of the work and the agents.

Who works on what, live
In use today
How many hours each agent works, waits or stalls, every day
In use today
AI use by each person’s agents, against a weekly budget
Partly built
Cost per task, in money
Partly built: on one computer
Alerts on wasted tokens and on personal use
Coming
Keys that no person or model ever sees, and one switch that stops every agent
In use today

Tokens are the units that AI use is billed in.

The rule we build to, for organisations: agents are measured closely, people gently. Each person sees their own data first, at team level, with no ranking of people.

The cockpit’s work screen: number tiles for what got done today, how long a change waits, checks passing and check minutes
Work in the cockpit: what got done today, how long a change waits, and the checks. Real numbers.

Connectors

  • Partly built: our own systems
  • Coming: your systems and sensors

Agents read from the systems an organisation already uses, and act in them only with approval.

  • Connected today, for our own company: code and tasks (GitHub), the game platform (Roblox), community chat (Discord), the shared calendar, phone notifications and this website.

Next Your systems, through one connector layer that starts read-only, and sensors in the field.

Built on the platform

The AI art line

A design agent, steered by our artist, makes 3D models, textures and interface art. A machine checks each delivery against a rulebook in seconds, one command puts it into the game, and a daily check makes sure every asset still works. Art never waits for approval: the artist asks for changes after it is in.

  1. Design agent
  2. Automatic check
  3. One command
  4. In the game
  • In use today: 239 assets in our game
  • Coming: for other studios and engines

Bring your organisation onto its own mycelium.

Tell us how your teams work today and where AI agents already help. A person on our team answers, in English or Hebrew.

info@myceliumlabs.com