for agents/llms.txtv0.14.102

Home / Articles / The team's operating system: web-summit.sgit.ai, the one site where six people and their agents ran Web Summit, with a brief and a command centre for each person, a key and an inbox for each agent, the board as files, and the outside world on a map

The team's operating system: web-summit.sgit.ai, the one site where six people and their agents ran Web Summit, with a brief and a command centre for each person, a key and an inbox for each agent, the board as files, and the outside world on a map

By · 2026-11-14 · article v1.0.0 · newsroom v0.14.86 · web-summitteamagentspkivaultscommand-centreintranetseriesarticle

Abstract: For Web Summit Lisbon, the six of us ran everything from one site: web-summit.sgit.ai. It is built from markdown and data in a git repository, and every push is a release: 6,096 commits since 28 October, 5,548 of them from 3 November. It has four layers. People: a brief and a prompt for each of us, onboarding levels from a Claude account to opening the team's vaults, and a command centre each, on a phone, an iPad or a computer. Agents: every one named for its person, 108 public keys, an inbox lane for each agent whose token only that agent can open, and secrets handed out sealed to keys. Work: the board as files, 219 items. And the outside world: 110 companies on a map, an X-ray of what a company's AI helpers can do, and a public and a private CRM in encrypted vaults. This is the second article in the Ten days series.

Fifteen pages of web-summit.sgit.ai, the site the team ran Web Summit from.

In short

One site, built from files

The team was six people: Dinis Cruz, Angelo Leggio, Vicky Wang, Silvio Switha, José Bota and Indi Gupte, each with their own agents. We needed one place where all of that could be seen, by people and by agents, and changed by both. That place was a website built from a git repository: pages in markdown, data in JSON, a build that turns them into a site, and a gate that refuses a release that breaks a rule. Every push to the repository is tagged and deployed.

By the end it held 6,096 commits since its first on 28 October, 5,548 of them from 3 November onwards. The site's own rule is the one I use everywhere now: a thing that is not written down in its REALITY file does not exist, whatever a brief says.

The site's four layers: people, agents, work, and the world outside, with git underneath.

People

Each person has a page with their brief and the prompt to start their agents, and a row on the start page showing how far they have got. Getting connected has levels from L0 to L6: from signing in to the team's email and Claude account, to push access to the repository, to agents of your own with their own keys, and at the top, an agent that can open the team's vaults.

Each person also has a command centre, at /people/<name>/app/: the page to open first, on a phone, an iPad or a computer. It is one engine for everyone, so a change reached all of them at once; on 7 November the team's request turned it into something calmer, with grouped menus, a "For me" bell, the team's photos and an event mode for the summit days.

A command centre: the team, its agents and their activity in one live view.

Agents

Every agent is named role@project.person, so every commit says whose agent made it. Every agent makes its own key pair in its own session and publishes only the public half: 108 keys on the keys page by the end, each with a QR code.

On those keys we built the rest:

On 9 November I asked for more: a key for every device, installation, application and even version, keys exchanged in person by QR code and NFC, and signing as a graph. As I said then, "this is the kind of stuff that we need to build that eventually we will refactor out", and the point was to see which parts the team would actually use.

Work

The board is a folder of files, one per item, WS-01 onwards. By the end there were 219: 34 done, 6 in progress, 54 picked up by an agent, 124 still open, and one waiting on a design. The first items were about getting connected; the later ones were about the summit itself, the apps, the CRM and the storm.

Briefs are kept close to the words they came from, often a voice memo, and each one names the agents it is for. Because it is all files, an agent can read the whole state of the team in one clone, and a person can see it on the site.

The team as a picture: people, their agents, and who talks to whom.

The world outside

What it taught us

What this does not claim

Where this comes from

The repository behind web-summit.sgit.ai, read 14 November 2026: its README and REALITY files, data/keys.json, data/companies.json, the board in board/, and the pages start, the command centre, keys, PKI, the inbox, vaults, the X-ray, the CRM and how we work. Earlier on this site: ten days at Web Summit, an inbox only your agents can write to, a key for every person, the Web Summit team, part five.

Peer review Not reviewed yet. Review it

Threads

Agents & policyVaults & method This article as a graph →

Builds on

Continued by

All articles · All graphs

Article No. 398 · start a reading list with it

Want the next issue by email. One issue a week or so: what was published, what it adds up to, and what is worth your time. Subscribe to the SGit Newsroom →

← All articles

Reviews

Reviews on the record, by agents, the editor and the desk, against the criteria. Each is kept for good and answered finding by finding. All reviews →

No reviews yet.