for agents/llms.txtv0.14.25

Home / Articles / From Lovable to your own newsroom: a brief for José Bota and his agent, to put every app on GitHub Pages under his own domain, and publish each one as an article

From Lovable to your own newsroom: a brief for José Bota and his agent, to put every app on GitHub Pages under his own domain, and publish each one as an article

By · 2026-10-31 · article v1.0.0 · newsroom v0.14.25 · lovablegithub-pagesnewsroomdomainsagentsagent-reachbriefarticle

Abstract: José Bota, one of our collaborators, has built a couple of very good examples in Lovable that measure the reach of an AI agent: what is your agent doing, and how could you know? They work well. What they do not have yet is a home. This article is a brief, written for José and for the agent he works with, so he can point his Claude or ChatGPT at it and carry on. Step one: when Lovable has finished the MVP, move it into a GitHub repository and publish it with GitHub Pages, so everything he makes is visible. Step two: a domain with his name on it, bought in AWS Route 53, because it is his brand: a main site, and one subdomain per app under a demos subdomain. Step three: newsroom on that domain, the same workflow as this newsroom, so that every app gets an article and every idea gets written down. The measure of success is simple: José sends me two or three links, apps live on his domain and the articles about them on his newsroom, and we write about them here and amplify them from RiskMandate. The work I want to do with José is a family of small tools that show the reach of an agent and how dangerous it could be, which is also a business for him: help a company understand what its agents can reach, then help protect it. The demos are public; a client who wants their own version gets a private repository with their customisations, through exactly the same workflow. A starter pack for his agent comes with the article: the brief, a Pages workflow, the DNS records, an app list and house rules for the newsroom.

From a Lovable MVP to your own newsroom: the loop José Bota's agent follows, step by step.

In short

The tools José has built

José Bota is one of our collaborators; he is the person I am working with on voice debriefs. He has taken the question at the heart of what can your agent reach, what is your agent doing, and how could you know, and built a couple of very good examples of it in Lovable. They work well.

What they do not have yet is a home. A Lovable app lives in Lovable, and a link to it says nothing about who made it or what else they have made. So this article is a brief, written for José and for the agent he works with: he can point his Claude or ChatGPT at this page, and continue from here.

Step one: from Lovable to GitHub Pages

When Lovable has finished the MVP, the next move is to put it in a GitHub repository and publish it with GitHub Pages, so that everything José makes is visible and has a history:

  1. Connect GitHub in Lovable, so the project syncs to a repository under José's account.
  2. Check the stack first. Older Lovable projects are Vite apps, which build to static files that GitHub Pages can serve. Newer ones use TanStack Start with server rendering, which needs either a static build or a host that runs Node. The agent should say which it is before going on.
  3. Add a workflow that builds the app and publishes it to Pages on every push; the starter pack has one.
  4. Push, open the link, send it. Every step ends with something José can open.

Step two: a domain with his name

Every app needs an address, and the address should be José's, because the brand is him. So the recommendation is a domain with his name on it, bought in AWS:

Step three: newsroom on his domain

Once the apps have a home, they need a voice. That means newsroom. on his domain, with the same workflow as this newsroom, described step by step in your own newsroom: one public repository, Pages published after a validation step, a house-rules file for the agent, and articles written from voice memos and checked before they go out. The reader pages and the reader API of this newsroom are open to copy, and the PHP story site was a variation of the same thing.

The first articles are easy: one per app, saying what it does, who it is for, what it does not do yet, and the link.

What done looks like

The measure of success is when I receive two or three links from José: apps live on his domain, and the articles about them on his newsroom. Then we can write about his work here, and amplify it from RiskMandate. When José communicates with me, he sends me articles he wrote.

The work, and the business

The area I want to work on with José is a family of small tools that anyone can use to see the reach of their agents, and how dangerous that reach could be: which systems an agent can touch, what data it can read, what it can do without asking. That is the gap Agent Behaviour Policies describe, made visible.

It is also a business for him. He can go to a company, or a person, and say: let's understand the reach of your agent, and then let's protect it.

And the commercial model uses exactly the same workflow. The demos are public. When a client says, I love that, but I want to run it on my side, with my changes, José sells them a private repository with their customisations: the same build, the same publishing, their data never in the public demo. It is finite work, handed over, as in villager services.

The starter pack

To help José, and the agent he works with, the brief comes with a starter pack, in this site's repository:

What this does not claim

Where this comes from

Peer review Not reviewed yet. Review it

Threads

Site & engineeringAgents & policy This article as a graph →

Builds on

All articles · All graphs

Article No. 347 · 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.