Before you continue...

CARLOS is an AI-enabled project, and all CARLOS software to date has been entirely produced using a combination of LLM-based coding products from Anthropic, OpenAI, and open-weight alternatives. Whilst CARLOS represents an approach to apply these products toward new standards of openness and release of corporate ownership of internet-based software and data, we realise that there are folks for whom use of these products is unacceptable. With that said, if folks will continue to use AI to produce software, our mission is to show that AI unlocks ways to produce better software for everyone.

One of the principles behind the project is full disclosure of AI use, and AI-written language, starting here.

Get started

Empty directory to a live URL.

Two routes. One makes every decision for you. The other walks you through the four that matter. Both are skills: you read this page, your agent fetches the file.

馃 That's factor X for AI-slop lines you'll spot across this site.

What you're starting with

Extraction is real and unfinished. Worth saying before the first command.

Rastrillo, the web framework that gives a CARLOS app its shape, is adoptable today: filesystem routing, manifests, a component vocabulary, a dev watch loop, and a bootstrap that speaks the platform's activation contract. The platform runs the fleet on the apps page. The carlos CLI ships, promotes, deploys and canaries. The platform's own repo isn't public yet; the framework, the skills and this site are.

Two routes

The recipe

Empty directory to a live URL, every decision made: rastrillo, server-rendered HTML, carlos deploy. For your first app.

SKILL 路 getting-started Read the skill

The menu

The principles and the four decisions: trust model, app shape, hosting, identity, with each source app's reasoning. For an app you have opinions about, or one you're bringing across.

SKILL 路 building-carlos-apps Read the skill

Give your agent the skill

Published at a well-known URL, digest-verified. Any agent can fetch them. Claude Code installs them as a plugin.

The index is at /.well-known/agent-skills/index.json, per the Agent Skills Discovery draft. As a Claude Code plugin they're carlos:getting-started, carlos:building-carlos-apps and carlos:delegate, the last being operator shorthand for splitting a job across tiered subagents with a reviewer beside anything risky. Source and install instructions are in the skills repository.

Then, in order

  1. Pick a trust class. Start with the data you handle and declare what the host is left holding. The classes.
  2. Build the app. The skill carries the procedure; rastrillo carries the shape. rastrillo.org.
  3. Ship it. carlos deploy builds, ships, promotes and waits for the live version. The CLI reference.
  4. Check it. Confirm the URL answers with an X-Carlos-Version header, and that the class you declared is the class you shipped. How the platform works.

Or have someone run it for you

Carloku is the flagship deployment, hosted. Same architecture, same guarantees. Don't be a hero!

Go to carloku.com Or run the platform yourself