# Docs for AI agents (/ai-agents)



This site is built to be read by models as well as people. Use this page to
point any agent (Claude, ChatGPT, Cursor, Codex, or your own) at the right
resources so it can act on the platform without guessing.

## Give an agent this site [#give-an-agent-this-site]

| Resource   | URL                                | Use it for                                                                                                       |
| ---------- | ---------------------------------- | ---------------------------------------------------------------------------------------------------------------- |
| Index      | [`/llms.txt`](/llms.txt)           | A map of every page with one-line descriptions, preceded by the platform facts below. Start here.                |
| Everything | [`/llms-full.txt`](/llms-full.txt) | All pages in one Markdown file, for large-context models or for building an index.                               |
| One page   | `/llms.mdx/<path>/content.md`      | The Markdown behind any page. Example: [`/llms.mdx/cli/commands/content.md`](/llms.mdx/cli/commands/content.md). |

Every page also has two buttons at the top right:

* **Copy for AI** copies the page as Markdown and appends the platform context
  block shown below, so the agent knows where the app, the CLI, and the rest of
  the docs live without a second lookup. Paste it into any chat.
* **Open in** opens the page in Claude, ChatGPT, or Cursor with a prompt that
  points at the Markdown endpoint, or copies just the page.

## Install the agent plugin [#install-the-agent-plugin]

For agents that work on your machine (Claude Code, Cursor, Codex), install the
skills plugin. It bundles the `acc` reference and task skills: first-time setup,
deploy a Docker app, deploy from a template, deploy a server, troubleshoot a
deployment.

```bash
git clone https://github.com/alternatefutures/alternate-clouds-skills ~/.alternate-skills
bash ~/.alternate-skills/install.sh
```

The installer detects which agents you have and links each skill into place.

## Let an agent run the CLI [#let-an-agent-run-the-cli]

`acc` works without a terminal prompt when it has a token and a project:

```bash
export AF_TOKEN="<personal access token>"   # acc pat create --name my-agent
export AF_PROJECT_ID="<project id>"         # acc projects list
acc whoami --json                           # pre-flight: exits non-zero when not signed in
acc services create --kind docker --image nginx:latest --port 80 -y
```

* Add `-y` to any command that would ask for confirmation.
* Use `--json` where a command offers it for machine-readable output.
* Without a TTY, commands that would prompt exit non-zero instead of hanging.
* Give each agent its own token so you can revoke it on its own. See
  [Access tokens](/guides/api-keys).
* `acc <command> --help` prints the options for any command. The full list is in
  the [command reference](/cli/commands).

## Platform context block [#platform-context-block]

This is the text **Copy for AI** appends to every page. Put it in a system
prompt, a `CLAUDE.md`, or an `AGENTS.md` when you want an agent to know the
platform before it reads anything else.

<AgentContext />

## What is not machine-operable yet [#what-is-not-machine-operable-yet]

* Signing up and adding a payment method happen in the web app.
* Composite (multi-service) templates deploy from the web app only.
* Pages in the Archive section describe the retired `af` CLI. Agents should
  not run those commands.