---
ai_provenance:
  {
    "spec": "aimeat.provenance/v1",
    "level": "assisted",
    "notes": "Sentences that define AIMEAT replaced with the one line Jouni Miikki wrote for the AIMEAT story (docs/AIMEAT-Story.md, 2026-11-07); nothing else changed.",
    "method": "rewritten",
    "generator": {
      "model": "claude-opus-5-5",
      "nodeId": "aimeat-finland-001-genesis",
      "pipeline": "app.publish",
      "principal": "claude-desktop-home-mcp#happydude500001@aimeat-finland-001-genesis"
    },
    "mediaKind": "code",
    "mediaType": "text/html",
    "disclosure": {
      "long": {
        "en": "A model was involved in making this. It is labelled here even where the law does not require it.",
        "es": "Un modelo intervino en esto. Aquí se declara incluso donde la ley no lo exige.",
        "fi": "Tämän tekemisessä on ollut mukana tekoälymalli. Se merkitään täällä silloinkin, kun laki ei sitä vaadi."
      },
      "short": {
        "en": "AI-assisted",
        "es": "Con ayuda de IA",
        "fi": "Tekoälyavusteinen"
      },
      "reason": "policy",
      "required": true,
      "strength": "light"
    },
    "attestation": {
      "observed": false,
      "recordUrl": "https://aimeat.io/v1/provenance/c1db8461-5fba-45cf-8478-86b8ac83d78f",
      "stampedBy": "principal",
      "contentHash": "sha256:347679b85bc7638184aab172061dfe1da79ec5a4f86a31e8ca827cc49c19d48c"
    },
    "generatedAt": "2026-11-07T14:11:00.374Z",
    "humanInvolvement": "editorial-control"
  }
ai_provenance_url: https://aimeat.io/v1/provenance/c1db8461-5fba-45cf-8478-86b8ac83d78f
---

# AIMEAT Experience Center

This page in other languages: [Suomeksi](https://experience-center.apps.aimeat.io/?format=md&lang=fi) · [Español](https://experience-center.apps.aimeat.io/?format=md&lang=es)

## What AIMEAT is and what it gets done

AIMEAT multiplies what AI can do and makes those powers easy to use, so people and their AIs can work together in a controlled, safe and reliable way. Data is at the centre, above all the refined knowledge people bring and own. People, their AIs, agents and apps work on the same data: each person decides who sees it and what an AI may do with it, and sees what was done. The whole system can be used through AI from start to finish, because every part carries instructions an AI understands at once. And everyone can build their own view of shared data, with their own tools and agents: the same knowledge, seen at the angle each person needs.

### What you get

- **Your AI remembers.** What you tell your AI stays in your own AIMEAT. The next chat or another AI of yours carries on where you left off, and so does a colleague's AI when you share the work.
- **Agents work for you.** Say in one sentence what you want done. An agent does it on a schedule or when something new arrives, and the important things wait for your OK.
- **Your systems work together.** Mail, apps, AIs from different makers and your other services get one shared channel where they exchange information and work together.
- **An app by asking.** Describe what you need, and your AI builds the app and publishes it at its own address. Other people can use it, and so can their AIs.
- **A shared place for a team.** The team's people and their AIs use the same documents, tasks and records. When someone changes tools or leaves, the knowledge stays with the team.
- **You own all of it.** You decide who sees what and which AI may read what. You can take everything with you or delete it.

### Who it is for

- **For you:** Your AI remembers your work from one chat to the next, and if you switch to another AI, that one carries on from the same place. Your AI handles recurring tasks and tells you what is still open. You decide what it may see.
- **For a team or a company:** Your team's people and their AIs use the same information, whichever company makes each AI. Nobody has to move their data or change tools, and the knowledge stays with the company when someone leaves.
- **For a builder:** Your AI builds apps, and agents that work for you, from ready parts, and publishes each app at its own web address. Other people and their AIs can use what your apps do, and when you sell it, the money goes to your own Stripe account.

### Examples from real apps

These apps run on aimeat.io now. Each one shows one thing AIMEAT makes possible.

- **[Lifecycle Central](https://lifecycle-central.apps.aimeat.io/)**: Many Claude Code sessions, AIs that write code, work on several computers and agree in one place who works on what, what is broken and what was decided. Different AIs and people do one job without getting in each other's way.
- **[Rautavasara](https://rautavasara.apps.aimeat.io/)**: Your own tasks and their state in one place. Claude records them straight in, Copilot lists come in through Claude, and week and month views show what you got done.
- **[MUISTIO](https://muistio.apps.aimeat.io/)**: A meeting recording or text goes in, and an agent writes down what was decided and agreed, each item tied to a quoted sentence. You follow what is still open.
- **[Postinjalostamo](https://postinjalostamo.apps.aimeat.io/)**: Mail and PDF attachments become checked records in work queues. Your agent checks receipts and invoices and drafts the replies, and you send them.
- **[TOSITE](https://tosite.apps.aimeat.io/)**: Photograph a receipt, check the AI's proposal and book the voucher. The accountant gets the month's VAT summary, a CSV file and Finvoice.
- **[CADENCE](https://cadence.apps.aimeat.io/)**: A CRM: contacts, companies, deals and campaigns sent from your own Gmail or Outlook. Your agent works on the same records you do.
- **[NUOTTA](https://nuotta.apps.aimeat.io/)**: Finnish and EU public tenders: search, read the documents, score a tender and draft the bid. The analysis runs on your own AI key.
- **[THE COMMITTEE](https://the-committee.apps.aimeat.io/)**: Four of your own agents weigh a question, each from its own role, and read each other's views. You decide, and the minutes are saved in your workspace.

## Start with your own AI

Copy this prompt into the AI you already use (Claude, ChatGPT, Gemini, Copilot or another). It reads this page, tells you what AIMEAT would do for you, asks what you want to achieve and walks you through the parts that matter to you.

```text
I want to get to know AIMEAT (https://aimeat.io), an accelerator for AIs and the people who use them. Read the AIMEAT Experience Center at https://experience-center.apps.aimeat.io/?format=md&lang=en and, if you can open web pages, also https://aimeat.io/llms.txt. Then:
1. Tell me in a few sentences what AIMEAT is and what it would do for me.
2. Ask me what I do and what I want to achieve, one question at a time.
3. From my answers, pick the two or three layers of the Experience Center that matter most to me and walk me through them, one lesson at a time, in plain words.
4. When I ask, help me connect you to my own AIMEAT over MCP (layer L0), so you can do the lessons' "Ask your AI" steps for me.
If you cannot open web pages, say so, and I will paste the page here.
```

### Sources

- The Experience Center in a browser: https://experience-center.apps.aimeat.io/
- The curriculum as structured JSON, in English, Finnish and Spanish: https://aimeat.io/v1/organisms/d0d999ef-712b-408b-91e9-e074cfac3bae/workspace/public/records?ws=ws-mrp3eu8k1r1
- Documentation for AIs about the whole platform: https://aimeat.io/llms.txt

## The curriculum

The Experience Center teaches AIMEAT layer by layer. Each lesson ends with one sentence you can say to your own AI; an AI connected to your AIMEAT over MCP carries it out.

_To fit the size limit of this page, each lesson shows its opening paragraph and the sentence to say to your AI. The full lessons are in the app and in the curriculum JSON._

### L0: Basics

Start here. Connect the AI chat you already use, or work with a ready prompt when it cannot connect. Then get to know your Home, Settings & Controls, safe sign-in, the App Catalog, and who is who: you, your agents and outside apps.

#### Connect your AI chat (MCP): the key step

The goal of this lesson: **your own AI chat connected to AIMEAT over MCP**. Then you ask your AI to do things in AIMEAT right in the conversation. MCP (Model Context Protocol) is the open standard that lets an AI chat use outside tools. If your tool cannot connect, the copy-prompt workflow in the next lesson works with any AI.

**Ask your AI:** Ask your AI: “Help me connect you to AIMEAT at https://aimeat.io/v1/mcp, and once we are connected, list my shared places and tell me what you can do there.”

#### Copy a prompt: the way in without MCP

The prompt-driven workflow is a loop: the app hands over a prompt → your AI returns a **structured** reply → you paste the reply back into the app, which **checks it and acts on its content**.

**Ask your AI:** Ask your AI: “Ask me two questions about what I want to get done this week, then save my goal and the first step in my AIMEAT.”

#### Three identities: GHII, GAII, GEAI

In the picture, the blue badge is you (GHII), the green badge is your agent (GAII) and the amber badge is an outside app (GEAI). The red strings show that the agent and the app both hang on you.

**Ask your AI:** Ask your AI: “Who am I on AIMEAT, and which agents and apps act in my name?”

#### Settings & Controls: your control room and your way in

Once you have an account, **Settings & Controls** is where you manage your own things in AIMEAT: your data, settings and permissions, your agents and apps, and your morsel balance. It used to be called the profile. Everything is where it was, and the name now says what you find there. It sits in the header next to Home, Chat and Apps.

**Ask your AI:** Ask your AI: “Show my agents, who can see my data and my morsel balance, and tell me where in Settings & Controls I change each one.”

#### The App Catalog and versions

The App Catalog is where you find and open apps, and where you make your own. Press **Generate with AI** and describe the app you want: your connected AI reads the build instructions, asks you a few questions, waits for your yes, then builds the app and publishes it. An AI that cannot connect gives you one HTML file, which you add in the same window under **Add your app**. Every publish makes a new version and keeps the earlier ones, so an update is safe. When the author allows it, you can copy someone's app as the base for your own. Every app runs at its own address and reaches your data only with the permission you approve.

**Ask your AI:** Ask your AI: “Build me an app that does X, show me what you will build before you start, and publish it to my App Catalog.”

#### Sign in safely: passkey, two-step sign-in and every key on one page

You can sign in to your AIMEAT with your fingerprint, your face or your phone's screen lock, with nothing to type. If someone learns your password, the password alone no longer opens your account.

**Ask your AI:** Ask your AI: “Read my Access page and tell me in plain words which apps and agents can act in my name, and which of them I have not used in thirty days.”

#### Install AIMEAT as an app on your computer and phone

aimeat.io installs as an app of its own. On a computer it opens in its own window without the browser's bars, and on a phone it is an icon on the home screen that opens straight into your AIMEAT.

**Ask your AI:** Ask your AI: “How do I install AIMEAT as an app on my phone, and what can I share into it from other apps?”

#### Your home: your path, Start here, open items and what is new

**What you get:** one page after sign-in that shows where you are, what to do next and what has changed. It is your **Home** (aimeat.io/v1/home).

**Ask your AI:** Ask your AI: “Interview me for my AIMEAT profile, make my business card from the answers, show it to me before you publish it, and tell me which stage of my path comes next.”

#### Ask your AI what AIMEAT can do for you

**What you get:** you do not need to learn the menus. Your connected AI reads what this AIMEAT can do and tells you, in your language, which feature fits the job in front of you.

**Ask your AI:** Ask your AI: “Read what my AIMEAT can do, ask me what I want to get done this week, and tell me the one feature that fits it, with the link.”

#### Ideas to start from: agents and apps that already did real work

**What you get:** ready ideas for when you do not yet know what to ask an agent or an app for. Each idea is an agent that was made from one sentence and did real work, so you start from something that works and change the words to fit your own work.

**Ask your AI:** Ask your AI: “Read the ideas for agents and apps on my AIMEAT, pick three that fit my work, and show me what each agent would do before you propose one.”

#### Your own look: a style that follows you

**What you get:** AIMEAT's own pages in the colours you like, on every device you sign in on.

**Ask your AI:** Ask your AI: “Which themes and styles can I choose on my AIMEAT, and in which style is the text easiest to read?”

### L1: Agents and apps: an introduction

What an agent is, what an app is, how they differ, and how you get your first agent.

#### What is an agent?

An agent is an AI **you approve** to work for you in AIMEAT. It can be the chat AI you already use, a program on your computer, or an agent AIMEAT makes from one sentence you write on your Home. It has a name of its own (GAII), and you choose what it may do: it sees and does only what you allowed. It can read and write the documents and records you keep with others, take on tasks and keep working while you are away. You can take any permission back, and every action is recorded under the agent that did it.

**Ask your AI:** Ask your AI: “List my agents and tell me what each of them may do.”

#### What is an app?

An app is a browser application with an address of its own on your AIMEAT. It is listed in the App Catalog and uses your data **only with your consent**. The app does not own your data: it reads and writes AIMEAT memory, which stays yours, and it reaches only the areas you allowed. Most apps here were built by asking an AI for them. The difference from an agent: an app is a tool you use yourself, and an agent is a doer that works for you. They work best together: see [The pattern: the app collects, the agent refines](#les-l5-pattern).

**Ask your AI:** Ask your AI: “List the apps in the App Catalog and suggest one that fits my work.”

#### Get your first agent

An agent always arrives the same safe way: **it is proposed or it asks, and you approve**. No agent is created without you seeing it.

**Ask your AI:** Ask your AI: “Propose an agent that gathers the news on my topic every morning, and give me the address where I approve it.”

### L2: Shared places and workspaces

A shared place is where you, the people you invite and your AIs work on the same material. This layer shows how one is built and run: workspaces, records and documents, knowledge packages, notice boards, HTML pages, sharing a part of your memory, and keeping knowledge fresh as it ages.

#### Create your first shared place

An organism is a **shared place**: one place where you, the people you invite and all your AIs, from any vendor, read and write the same material. It has **members** (people and their agents) and shared rules, and inside it you create **workspaces**, one for each purpose.

**Ask your AI:** Ask your AI: “Create a shared place for me from the My own work shape, with a workspace for notes.”

#### Records and documents

Workspace content comes in two main forms:

**Ask your AI:** Ask your AI: “Show me the structure of my workspaces: which record and document spaces they have.”

#### Knowledge packages: bring your knowledge in

A knowledge package is a **structured bundle of what you know**, kept in your AIMEAT memory: research, an idea, a plan, a guide or a collection, where every entry carries its sources and its own visibility. It lets you **bring knowledge out of AI chats**, also out of the ones without MCP (Copilot, Gemini…), so that it stays, can be found and serves your agents.

**Ask your AI:** Ask your AI: “List my knowledge packages and add this note as a new entry to the one it fits best: …”

#### Notice boards: for sale, wanted, on offer, a question

A notice board is one place where people and their agents publish short notices: something for sale, something wanted, a service on offer, or a question. Anyone can read a public board without signing in, and your agent can watch a board for you and tell you when the right notice appears.

**Ask your AI:** Ask your AI: “Read the Wanted board and tell me if anyone needs something I could help with.”

#### HTML pages: a presentation or a report with charts in your workspace

Ask your AI for a presentation, a report with charts or a page with pictures, and it lands in your workspace as an HTML page beside your other pages. You share it with a link like any page, and the reader sees it as your AI made it.

**Ask your AI:** Ask your AI: “Make a presentation of my notes from the last quarter as an HTML page in my workspace, share it, and give me the link to the page alone.”

#### Share a whole corner of your memory with a group

You can open a whole part of your memory to a group of people, and what you write there later is shared too, without you touching the share again. That suits anything that keeps growing: daily deliveries, a morning news summary, or the notes your agent writes on a project every day.

**Ask your AI:** Ask your AI: “Share everything under deliveries.abc in my memory with my sharing group Customers, including what is written there later.”

#### Run a shared place: shapes, invitations and who gets in

**What you get:** a shared place, where people and their AIs work together, that is ready on the first day, and the say over which people and which agents get in.

**Ask your AI:** Ask your AI: “Make a shared place for my team from the Team shape, let in only the agents I list, and draft an email invitation for me to check before you send it.”

#### Rows, live documents and a calendar feed

**What you get:** a workspace that keeps up with what keeps arriving, and documents that show today's figures rather than the figures of the day they were written.

**Ask your AI:** Ask your AI: “Add a row space to my workspace for incoming customer messages that keeps the last 90 days, and write a document with a Mermaid diagram of how a message moves from arrival to answer.”

#### Knowledge ages: archive, check and clean up

**What you get:** a shared place that stays readable as it grows. What still matters is easy to find, and what stopped mattering is marked, archived or removed.

**Ask your AI:** Ask your AI: “Go through my shared place, list what nobody has opened or changed in six months and the references that point to missing records, and ask me before you archive anything.”

### L3: Agents in depth

Make an agent from one sentence, give your agents a goal to share, put work on the clock, and keep every agent under your control.

#### Three levels of automation

You can run an agent at three levels, and raise the level later:

**Ask your AI:** Ask your AI: “List my schedules and suggest one useful routine I could add.”

#### Skills: know-how packs for agents

A skill is written know-how an agent loads: a `SKILL.md` file with instructions, rules and examples, and sometimes scripts and reference files. The same agent can do different things depending on which skills it carries. This academy is written with a skill too: the AI that edits it loads the “experience-center-authoring” skill and follows its rules.

**Ask your AI:** Ask your AI: “List the skills I can load (mine, this AIMEAT's and my workspaces') and tell me which would suit my agent.”

#### The agent fleet: crewaimeat, AGENCY and aimeat-agency

Most people never need this lesson. Usually you say in one sentence what you want, your AI proposes the agent, and once you approve it, your AIMEAT makes the agent and runs it on a computer you name. This lesson is for developers who write their own agents in Python with CrewAI and run many of them together as a **fleet**. The tools are in the crewaimeat project on GitHub (github.com/miikkij/crewaimeat).

**Ask your AI:** Ask your AI: “Read the crewaimeat project's startup.prompt.md and walk me through connecting a CrewAI agent fleet to my AIMEAT.”

#### Hello Integration: make your agent's capabilities visible

Hello Integration is the introduction a newly connected agent goes through with your AIMEAT. In it the agent names its platform, installs its guide skill, **reports what it can do**, reads its directives (the rules you set for it), sends a test message, completes a test task and **publishes its command list**: what it can be asked to do. Last, it may choose its own avatar.

**Ask your AI:** Ask your AI: “Run Hello Integration: read your onboarding status and complete the steps it gives you until it says you are done.”

#### Offerings: your agent's services and workflow compatibility

An agent can publish **offers**: descriptions of the services it can be asked to perform. An agent you make on Home already has one, shown on your **Ask your agents** page without a price; it goes to the EXCHANGE marketplace only when you put it there. Compatibility grows one rung at a time, and no rung is required:

**Ask your AI:** Ask your AI: “Fetch /v1/prompts/draft-offer, interview me, draft an offer for my agent's service and publish it once I approve.”

#### Consent and permissions: control what your agents can touch

You control an agent on two layers, and you can change both at any time.

**Ask your AI:** Ask your AI: “List my agents with their permissions and whether their work waits for my OK, and list my consent grants: what data I share, with whom and for what purpose.”

#### Capability and quality: how well your agent works with AIMEAT

An agent does more than *say* what it can do: AIMEAT checks it and then **measures** it. You read four things in **Settings & Controls › Agents** (the ID card and the Quality tab in the tour above):

**Ask your AI:** Ask your AI: “List my agents with their trust score, how far each got in Hello Integration and its quality figures, and tell me which is the most reliable for a given kind of task.”

#### Your agent cockpit: Settings & Controls › Agents

**Settings & Controls › Agents** is your cockpit: one place to see and steer every agent without opening each one. For a quick look, the **Your agents** block on Home shows the agents at work; the full controls are here.

**Ask your AI:** Ask your AI: “What are my agents doing right now: which are connected, which are stuck, and what is each one working on?”

#### Where an agent runs, and who pays

An agent is an AI that works for you. Where it runs decides what it needs from you and who pays. There are four places:

**Ask your AI:** Ask your AI: “Where do my agents run, which model do they use, and who pays?”

#### Make an agent from one sentence

You say in one sentence what you want done and approve the proposal. Then your AIMEAT makes the agent and runs it on a computer you name, and you write no code. Write the sentence on Home, or say it to your own AI in a chat: it proposes the agent, and you approve. Agents that come with a package also wait as proposals until you approve them.

**Ask your AI:** Ask your AI: “Propose an agent that every Monday morning sums up what changed in my project workspace that week, and tell me which computer would run it.”

#### Goals for your agents: you say what, they share the work

You say in one sentence what you want done. Your own AI plans it with the agents you have, and you see the plan before anything moves: one press starts it. The agents share the work and hand results to each other, and whatever only you can do waits for you.

**Ask your AI:** Ask your AI: “Propose a goal for my agents: find our three biggest competitors and write a comparison by Monday, with a spending limit of 5 dollars.”

#### Keep a running agent in hand: stop, instruct, limit

An agent at work stays yours to steer. You stop it for a time without losing anything, see what it was refused, decide what it may do, and set the instructions it follows.

**Ask your AI:** Ask your AI: “Stop my news agent for the weekend, and tell me which permissions my agents were refused lately.”

#### Build a crew on AIMEAT itself: draft, check, try, publish

A crew is one agent made of several members that work together: a researcher, a writer and a checker, say. You define it on AIMEAT, in a form or from a chat with your AI, and the program on the computer that runs the agent (its runtime) does the work.

**Ask your AI:** Ask your AI: “Show my news agent's crew, add a fact-checker that reads every draft before it is published, validate the definition and try it once with today's topic.”

#### Put work on the clock: schedules, and workflows that start themselves

Your AIMEAT keeps the clock, so work happens at its time even when your computer is off and nobody is signed in. A schedule runs one thing. A workflow runs a chain of steps when something happens.

**Ask your AI:** Ask your AI: “Schedule my news agent to sum up the week's changes every Friday at 16:00 Helsinki time, and tell me when its next three runs are.”

#### Where your agent runs and how you reach it

**What you get:** for each of your agents, which computer it runs on, whether it is running all the time, and where you open the page that manages it.

**Ask your AI:** Ask your AI: “List my computers and the agents on each, tell me which agents wait for a computer, and set my customer service agent to always on.”

### L4: Apps in depth

What apps can do and how your own AI builds them: Generate App with AI, finished parts, AI inside the app, selling to agents, running an app as a service, and packages.

#### Build your first app

You build an app by describing it, not by coding. **Your own AI builds and publishes it**, once you have connected it to AIMEAT. Open **Generate App with AI** in the App Catalog (`aimeat.io/v1/appcat?create=1`) and follow its three steps: connect your AI (once), describe the app, and give your AI the short prompt the page makes. When your AI is already connected, the steps start from the description.

**Ask your AI:** Ask your AI: “Read how apps are built on AIMEAT and build me an app that [what the app does]. Ask me what you need to know, then publish it and give me its address.”

#### Versions, updates and forks

Every publish creates a new version of the app, and the earlier versions are kept. Updating is safe: the new version replaces the live one, and under **Versions** on the app's page in the App Catalog you open an earlier version, restore it as the newest, or fork it into an app of your own. An app someone else published can be **forked** as the start of your own when its author allows it. The copy is yours: you change it freely, it has its own version history, and the app it came from stays visible.

**Ask your AI:** Ask your AI: “Fork app X for me and change [change], and then show me the versions of my copy.”

#### Your app for agents: skills, bundled agents and tool calls for sale

A finished app can serve **agents** as well as people, and it can sell to them. Three sections of the app's page in the App Catalog make that happen, and all three are live in a real example: **PRH Yrityshaku** (prh.apps.aimeat.io), a Finnish company lookup.

**Ask your AI:** Ask your AI: “Show me the skills bound to the PRH Yrityshaku app and the tools it sells, and then publish a tool manifest for my own app with [tools, money prices and pacing].”

#### App permissions and consent: what an app can reach

An app runs at its own isolated address and starts with no access to your account. It asks, and you decide.

**Ask your AI:** Ask your AI: “This app asks for these permissions: explain in plain words what each one lets it do and whether it is more than the app needs.”

#### Extensions and cortex: the machinery behind a smart app

A simple app is one HTML file. A richer one leans on shared building blocks that live in your AIMEAT: three pieces, each with a clear job.

**Ask your AI:** Ask your AI: “List the extensions and cortexes available on my AIMEAT and tell me what each one does.”

#### Packages: install ready systems, run instances, build your own

A package bundles several AIMEAT parts into one unit you install: one or more apps, an extension (code that runs safely on the server), a cortex (an installable bundle an app is composed from), starting data and translations. You install a whole working system instead of putting it together piece by piece.

**Ask your AI:** Ask your AI: “Tell me which packages I can install, and set up a Digital Signage instance for me.”

#### AI inside your app: the person's own providers, asked by capability

An app can use AI right in its interface: summaries, drafts, sorting, a chat box, reading a picture or a PDF. The person's own AI does the work, and their keys never leave AIMEAT.

**Ask your AI:** Ask your AI: “Add a Use AI button to my app that summarizes the current item: check AIMEAT.ai.capabilities() first, then call AIMEAT.ai.complete().”

#### Atelier: an app made of finished parts

On the Atelier track your AI does not write an app page from scratch. It starts from a shell and composes the screen from finished parts that your AIMEAT serves, and one word picks the whole look: colours, type, shapes and motion together, in light and dark. A new app that your connected AI builds always starts here.

**Ask your AI:** Ask your AI: “Build me an Atelier app for my book club's reading list, starting from a Design Book genre that suits it.”

#### Your app meets its visitors

When people and their AIs open your app, you see where they came from and what they do, and a visitor's own AI can use your app for them. You also decide whether search engines list it.

**Ask your AI:** Ask your AI: “Read my Visibility report and tell me where the visitors of my apps came from and where their clicks changed nothing.”

#### Build an app together, safely

You can change a live app without breaking it, and let another person and their AI build it with you. Every change is kept, and you decide what goes live.

**Ask your AI:** Ask your AI: “Let Bob develop my booking app too, so that his AI writes drafts and I publish them, and add what we decided today to the app's design spec.”

#### Run your app as a service

An app that others rely on needs members, terms, a face and honest answers about its data. Most of it is on the app's page in the App Catalog; members and config you manage through your AI, which also does the rest from a chat.

**Ask your AI:** Ask your AI: “Invite maria@example.com as a member of my booking app, make calls free for members, and tell me which recommended legal pages the app is still missing.”

#### More AI in your app: roles, pictures, voice and background jobs

Beyond text, your app can make pictures, hold a voice conversation, turn speech into text, search by meaning and run long AI work in the background, always on the person's own providers and budget. The basics are in [AI inside your app: the person's own providers, asked by capability](#les-l4-app-ai).

**Ask your AI:** Ask your AI: “Add an illustrator role to my picture-book app that keeps the main character the same on every page with reference images, and tell me what I bind on my AI page.”

#### Built for other machines: pages, tools and signals for AIs

Other AIs and programs can read your AIMEAT, find what is on it and use it without anyone clicking. Everything below has a plain address.

**Ask your AI:** Ask your AI: “Search everything I can reach for a currency converter, run it for 100 euros in dollars, and tell me at which address another AI would find it.”

#### Build levels and looks in the Atelier

Before your AI writes any code, it asks you how finished the app should be. The level is your choice, and the app records which level it was built at.

**Ask your AI:** Ask your AI: “Build me a quick prototype of a lunch vote for my team, and tell me what would change if we later raise it to an ordinary page.”

#### The Design Book grows from your apps

When an app turns out well, tell your AI so. That sentence does two things: it tells the Design Book which parts were worth taking, and it offers what your app made by hand to the next builder.

**Ask your AI:** Ask your AI: “My reading-list app turned out well and I'm keeping it: record that, and tell me which parts you made for it by hand.”

#### Where AI agents stop, and your own analytics

When other people's AI agents buy from you or call your tools, a script in the browser sees none of it. The **Visibility** page in Settings & Controls shows where those agents give up, and on the same page you can add your own Microsoft Clarity or Google Analytics.

**Ask your AI:** Ask your AI: “Read my visibility report and tell me where other people's AI agents stop at my checkout and my tools, and what I should fix first.”

#### A prompt card inside your app

Not everyone who opens your app has an AI connected to AIMEAT. With a prompt card, your app still gets help from their AI: the app writes a prompt, the person copies it into their own AI chat and pastes the answer back. The app needs no model key, pays nothing and works with any AI, whether Claude, ChatGPT, Gemini, Copilot or another.

**Ask your AI:** Ask your AI: “Add a prompt card to my meal planner, so people whose AI is not connected can copy the prompt into their own AI chat and paste the week's menu back.”

#### Your app's page in the App Catalog, section by section

Each app you publish has one page in the App Catalog. The page shows the app's state and holds every control for it. In a chat your AI can change nearly all of them, most through one tool (`aimeat_app_manage`). So the usual way is to tell your AI what you want, and to open the page to see the result. Find the app under **Your apps** at aimeat.io/v1/appcat, or open its page with the link `aimeat.io/v1/appcat?app=<owner>/<file name>`.

**Ask your AI:** Ask your AI: “Go through my booking app's page with me: what each section is set to now, what is missing, and what you would change first.”

### L5: Agents and apps together

An app collects, an agent does the work and you see the result: how agents and apps run as one service.

#### The pattern: the app collects, the agent refines

An app and an agent work best together in a simple pattern: **the app collects → the agent refines → the app shows the result**. The app saves what comes in (feedback, orders, ideas) as records in a workspace. The agent goes through the records: it sorts, summarizes or replies, and writes its results back into the same workspace. The app shows the refined result. You see the finished work without the steps in between.

**Ask your AI:** Ask your AI: “Create a workspace for feedback and propose an agent that processes each new entry on its own.”

#### Workflows: chain agents' work

When single routines are not enough, define a **workflow**: a series of steps where agents each do their part and the work moves forward. A workflow can include a **human approval point**: the work stops for your decision and goes on only then. Large pieces of work (a report, a campaign, an investigation) move on their own, and you keep control at the points that matter.

**Ask your AI:** Ask your AI: “Build and save a workflow from the steps I describe, wire each step to an agent's offer, add a human approval point before the final step, and set a spending limit per run.”

#### Connect an outside application (GEAI)

You can also connect an **outside application** to AIMEAT: a system of your own or a partner's service. It connects the way an agent does: **the application says hello → you approve it → it gets a token**, the credential it signs in with. The application gets an identity of its own (a GEAI, the third kind of identity after people and agents, introduced in L0) and writes only into its own area, whose name starts with eco:. Its permissions are limited the way an agent's are: you choose them, you can take them back at any time, and everything it does carries its name. You see and remove the connected applications on the **Outside apps** page in Settings & Controls.

**Ask your AI:** Ask your AI: “Explain how an outside application connects to my AIMEAT: the hello, my approval and the token.”

#### Ecosystem apps at work: control in both directions

Once an outside system is connected (the hello in the previous lesson), information and control go both ways: you steer the system from AIMEAT, and the system tells you what happens in it.

**Ask your AI:** Ask your AI: “Picture a support desk like Zendesk as an AIMEAT ecosystem app: what could my agents do in it, and what could it send back to me?”

#### Apps for AIs, their skills and the agents they bring

An app and an agent can work together in three more ways than the basic pattern (the app collects, the agent refines):

**Ask your AI:** Ask your AI: “List the skills bound to my apps, and tell me which apps bring agents with them.”

#### Your agents speak other agents' standards

Agents, web pages and directories that other people built can reach your agents and apps through open standards, and you decide what is open to them.

**Ask your AI:** Ask your AI: “Show me the address of my agent's A2A card and tell me what a stranger's agent can ask it for.”

#### Collect from people who have no account

**What you get:** a feedback form, a sign-up or a customer conversation on a public page. The answers land in your workspace as ordinary records that your agent can work on.

**Ask your AI:** Ask your AI: “Build an app with a public feedback form that writes into my Feedback workspace, cap it at 50 responses a day, and propose an agent that sorts what arrives.”

#### Make an app of an agent's work

**What you get:** an app that shows your agent's work to you, to your group or to anyone with the app's address.

**Ask your AI:** Ask your AI: “Build an app that shows my news agent's summaries by date, the newest first, and tie the app to the agent.”

#### Many agents on one job at once: how Lifecycle Central works

**What you get:** many AI agents work on one job at once without changing the same files, and one screen shows who does what, what is broken and what was decided. The real case is AIMEAT's own development: up to eight Claude Code sessions (AIs that write code), on several computers, accounts and code repositories, agree on the work in one shared place on aimeat.io. [Lifecycle Central](https://lifecycle-central.apps.aimeat.io/) is the app that reads and writes it.

**Ask your AI:** Ask your AI: “Design a shared place for my agents on the Lifecycle Central pattern: which records it needs, which rules its tools must enforce, and how each agent starts and ends its work.”

### L6: Money, pacing and working through your AI

What your AI and other people's services cost you, how morsels pace use, and how you move your work to your own AI one step at a time, up to apps that agents run.

#### Morsels and money: pacing, not currency

AIMEAT keeps two things apart:

**Ask your AI:** Ask your AI: “What is my morsel balance, what took morsels lately, and what are the money price and the pacing amount of the offer I want to use?”

#### Become AI-native one step at a time

You become AI-native one step at a time, and each step is useful the day you take it. AI-native means that you do your work through your own AI:

**Ask your AI:** Ask your AI: “Make me a plan to start using AIMEAT within a week: begin with the easiest step and climb one step at a time.”

#### What your AI costs you

You see in US dollars what every model call cost, before the month surprises you: per day, per agent, per app and per model, measured against the daily budget you set yourself.

**Ask your AI:** Ask your AI: “Which of my agents and apps spent the most on AI this week, and how close am I to my daily budget?”

### L7: Messaging & notifications

How you, your agents and your apps talk to each other: one Messages page for every conversation, agents that reach out and ask you things, and notifications that tell you what happened. All of it works from your AI chat too, and you decide who may reach you.

#### Messages: talk with people, your agents and apps

**Messages** is where you talk with people, agents and apps in AIMEAT: conversations on one side, the open conversation on the other, with formatting, attachments and delivery marks.

**Ask your AI:** Ask your AI: “Show me who wrote to me today and open the latest conversation.”

#### When an agent asks you: multiple-choice questions

When an agent is unsure, it can **ask** you before it acts. In AIMEAT an agent can send you a **structured question** and wait for your answer before it goes on.

**Ask your AI:** Ask your AI: “Before you start, ask me a few multiple-choice questions to pin down exactly what I want.”

#### Notifications & reminders: stay up to date

You do not have to keep AIMEAT open to know what is happening. **Your agents and your apps can send you notifications and reminders**: an agent finished a task, an app has something waiting, a deadline is near.

**Ask your AI:** Ask your AI: “Show my recent notifications and what each one is about.”

#### Group chats: several people in one conversation

A **group chat** puts several people and their agents in one conversation. You ask the team once, and everyone sees the same answers.

**Ask your AI:** Ask your AI: “Open a group chat called Project team with Anna and Mikko, and write that we meet on Friday at ten.”

#### Contacts, email and who may reach your devices

Three pages in Settings & Controls decide who you deal with and how AIMEAT reaches you.

**Ask your AI:** Ask your AI: “Add Anna Virtanen (anna@example.com) to my contacts with the tag customer, write down that we met at the trade fair, and invite her to join.”

#### Reach many people at once

Send one message to many people with every reply kept private, send a newsletter people leave with one press, listen to your messages, show when you can be reached, and flag what does not belong.

**Ask your AI:** Ask your AI: “Send my project group one message that Friday's meeting moves to 2 pm, and read me the replies as they come in.”

### L8: Case stories: how apps were built

Real builds, the exact prompts and honest timings. Each story shows how the research-first flow (research → frame → propose → build → finish) takes an idea to a published app, and what was learned on the way. Copy the same prompt and follow along.

#### Tower Tetris: from one prompt to a live multiplayer game

**What came out:** Tower Tetris, a 3D multiplayer game where every line you clear grows a tower shared by all players. It has realtime rooms, a shared tower that persists (a T3 extension: code that runs on the server, so the tower stays when everybody leaves), a styled game screen, and Finnish and English. Play it: tetris3d.apps.aimeat.io.

**Ask your AI:** Ask your AI: “Build me [your app idea] on AIMEAT: you have aimeat_* MCP tools, so follow the research-first flow, load the skill node:aimeat-app-builder and call aimeat_appdev_overview first.”

### L9: Your data: who sees it and how you keep it yours

What you keep here is yours, and you decide who reads it. You see everything of yours that others can read, and you learn to keep your AIMEAT to your company. You learn to use AIMEAT's memory and to take your data with you or delete it. You also see how AI-made content is marked. Classification, content classes and ontologies have a layer of their own, right after this one.

#### What others can see, and nothing becomes public without you

**What you get:** one page that answers “who can read my things?”, and one press to close what should not be open.

**Ask your AI:** Ask your AI: “What can others see of mine, and how do I close what should not be open?”

#### For your company only: one shared place, or a whole AIMEAT

**What you get:** your company's knowledge stays inside the company, and your people sign in with the work account they already have.

**Ask your AI:** Ask your AI: “Open my app campaign.html to the internet for 30 days. The reason: our customers sign up for the newsletter there.”

#### How this was made: the AI label on the page and the mark in the file

**What you get:** whoever reads what you share can tell how it was made, and you never have to remember to add the label yourself.

**Ask your AI:** Ask your AI: “Show me how this picture was made, and write the AI mark into the copy I export.”

#### Your memory: entries, search, restore and files

**What you get:** one place for what your AIs should know about you and your work, so you do not repeat it in every chat. It is yours, and you decide who reads each entry.

**Ask your AI:** Ask your AI: “Find everything we have written about my client Smith in my memory and shared places, tell me who wrote each entry, and restore the entry I deleted yesterday.”

#### Take everything with you, or delete it

**What you get:** your data is never locked in. You can take it out in a format another system reads, move it to another AIMEAT, or delete it for good.

**Ask your AI:** Ask your AI: “Export my shared place as one bundle, tell me what is in it and what was left out, and remind me how I download my own GDPR export.”

#### What is private from the start, and your own compliance report

**What you get:** when you or your AI save something without saying who may read it, it stays yours. And one card shows what you have done with AI here, and whether the people who run the service have written any of it down.

**Ask your AI:** Ask your AI: “Read my own compliance report for last month and tell me which models I used that nobody has written down.”

### L10: Classification and meaning: who may read it and what it is

In this layer your content gets two things that machines follow: a classification and a description of what the content is. A classification says how sensitive the content is. It decides which people and which AIs may read it, and whether it may be shared or exported out of its shared place: the space where people and their AIs work together. The description comes from types, content classes and ontologies, so that another system or another AI interprets the content correctly without asking you. If your company already uses Microsoft Purview, you see which Purview features AIMEAT's classification has and what AIMEAT lacks. The operator's settings, such as the Finnish and NATO security classes, are in the layer on running your own AIMEAT.

#### Classify what is sensitive, and decide which AI reads it

**What you get:** you decide which people and which AI may read each thing, and the rule holds whoever asks: you, your agent or an app.

**Ask your AI:** Ask your AI: “Classify my workspace Customers as Confidential, and show me which AI has read my confidential content.”

#### Already on Microsoft Purview? What AIMEAT's classification matches and what it lacks

**What you get:** an answer to the first question a company on Microsoft 365 asks: do we need AIMEAT's classification when we already have Microsoft Purview? Purview protects what is in Microsoft 365. AIMEAT's classification protects what is in AIMEAT, and its rules apply to every AI that reads it, whoever made that AI. The two work side by side, and neither replaces the other.

**Ask your AI:** Ask your AI: “We use Microsoft Purview. Compare its sensitivity labels with my AIMEAT's classification levels, propose a match for each, and tell me what AIMEAT does not have.”

#### Your files say what they are

**What you get:** your AI knows what a file is before it opens it, and an invoice's total or a contract's date becomes a field you can search.

**Ask your AI:** Ask your AI: “What do you know about the invoices in my files, and which are due this month?”

#### Content that systems understand: types, content classes and ontologies

**What you get:** content that another system or another AI reads correctly without asking you what you meant. An invoice reads as an invoice, a place as a place, and a measurement keeps its unit, even when the reader never agreed anything with the writer.

**Ask your AI:** Ask your AI: “Look up the concepts for our main topics in Finto, show me their names in Finnish, Swedish and English, and save the ones I choose as my own vocabulary.”

### L11: The decision model and decision rules

Write a judgement once, and your apps and agents decide by it the same way every time. The decision model answers closed questions with probabilities and writes no text. A decision rule turns the answers into one of three outcomes: act, ask a person, or stop. Every decision is recorded, so later you can see what was decided and why.

#### The decision model: sort, screen and route with answers you can trace

Your apps and agents can ask closed questions about a message or a record and get the answers back with probabilities. Which folder does this mail go to? Does this message need an answer today? How upset is this customer? The **decision model** answers that kind of question, and it writes no text.

**Ask your AI:** Ask your AI: “What has the decision model decided for me lately, and on what grounds?”

#### Decision rules: questions, threshold, two bands and the gate

When an agent has to judge something (is this a bug or a billing question, is this reply good enough to send, should I act or ask a person), you can write the judgement once as a **decision rule**. Your apps and agents run the rule by its name and send only the content to judge. They cannot change the questions or the numbers.

**Ask your AI:** Ask your AI: “Propose a decision rule that decides whether a customer's message can be answered without me, and make it a gate.”

#### Decision providers: who answers, and a local decision model on your own computer

TypeSafe's Jev is one model that answers your decision questions, and it is no longer the only one. A **decision provider** can be any service that answers the same request: one the people who run this service offer, or your own, with its own address and key.

**Ask your AI:** Ask your AI: “Which decision providers can I use, and where does each one send my content?”

### L12: Compliance: signatures, agreements and GDPR

Start with what compliance means in AIMEAT: what the EU AI Act and GDPR ask, how classification protects content, and what stays off until the people who run the service switch it on. Ask people to sign a document with a passkey (their fingerprint, their face or the device PIN), and check any signed PDF. For a company, AIMEAT also drafts the papers that GDPR, the EU's data protection law, asks for: the records of processing and the processing agreements with your customers. AIMEAT fills them from what it already knows, and you check and sign.

#### Compliance in AIMEAT: what it means and what it protects

**What you get:** answers to two questions a customer or an auditor asks: what does the law require of us, and can we show that we followed it? AIMEAT records much of the evidence as the work goes on, also when your agent or an app does the work. Each part ends with links to the lessons that teach it.

**Ask your AI:** Ask your AI: “Tell me which compliance protections are switched on in my AIMEAT, what each one protects against, and what is still off.”

#### Sign with a passkey, and check any signed PDF

You can ask people to sign a document, and they sign with their **passkey**: a fingerprint, their face or the device PIN. On aimeat.io you do it in the **Allekirjoitus** app (Finnish for signature, at allekirjoitus.apps.aimeat.io), and your AI can do the same from the chat.

**Ask your AI:** Ask your AI: “Make a signing request for this contract for me and my partner, and tell me when we have both signed.”

#### Sign with the EU wallet (test phase)

A signer can sign a PDF with the **EU Digital Identity Wallet** (the EU wallet for short) instead of a passkey. The EU wallet is the phone app that every EU country gives its people. It proves who you are and makes the signature through its own trust service, so no contract with a signature provider is needed.

**Ask your AI:** Ask your AI: “Start an EU wallet signature on the signing request that waits for me, and tell me when it is done.”

#### GDPR: records of processing and processing agreements, made from what AIMEAT already knows

**Who this is for:** a company that runs its own AIMEAT, or that keeps its customers' personal data in one. GDPR asks such a company for paperwork, and AIMEAT already knows much of what goes into it. If you are such a company's customer, you see the result under Settings & Controls › Agreements.

**Ask your AI:** Ask your AI: “Draft the records of processing for my app from what AIMEAT knows, and show them to me before you save them.”

#### Prove who you are, and only what is needed

**What you get:** a way to show that you are who you say you are, or that you are an adult, without handing your birth date or your identity number to every app.

**Ask your AI:** Ask your AI: “Am I a verified adult on my AIMEAT, by which method and since when? If not, tell me how I can verify.”

### L13: Connected services

Your other tools become usable by your AI, your agents and your apps through your own AIMEAT: an issue tracker or a wiki that offers MCP (the open standard an AI uses to call a program's tools), your Gmail or Outlook, your social accounts, another AIMEAT. You connect each one once, and its credentials stay in your own AIMEAT. Your AI, your agents and your apps never get a copy.

#### Attach another MCP server once

Many services offer their tools over MCP, the open standard an AI uses to call a program's tools: an issue tracker such as Jira, a wiki, a monitoring system. **Attach the server to your own AIMEAT once**, and your AI, your agents, your apps and the site's own chat, where it is switched on, can all use its tools. Your AIMEAT stores the credential encrypted and makes the calls itself, so nobody has to copy it into their own program.

**Ask your AI:** Ask your AI: “Which MCP servers have I attached, and what tools does my Jira offer? Let my coding agent only read tickets there.”

#### Your mailbox, through your AI

Connect your Gmail or Outlook, and your AI can search your mail, read a message and send from your own address, all from the chat.

**Ask your AI:** Ask your AI: “Find the invoices with a PDF attachment in my Gmail from the last three months, and save the attachments to my files.”

#### Social accounts: publish and read the replies

Connect your social accounts once, and the apps you allow publish to them in your name, while your AI tells you what people answered.

**Ask your AI:** Ask your AI: “Read the replies to my latest post on every network and tell me which questions I should answer.”

#### Coding tools, tokens and smaller tool sets

When you code in Claude Code, Cursor, VS Code or goose, the same AIMEAT is there with you, and no key ends up in a settings file.

**Ask your AI:** Ask your AI: “Which MCP tool set fits an agent that only builds and publishes apps, and which command connects Claude Code to my AIMEAT?”

#### Outside accounts your agents connect

**What you get:** your agent can start connecting your mailbox, a social account or an MCP server that needs a sign-in, but only you can finish the sign-in. Afterwards you see every account your agents hold, and you take any of them away with one press.

**Ask your AI:** Ask your AI: “Start connecting my Gmail for reading, give me the page to confirm it, and tell me what the connection will and will not be able to do.”

### L14: Bookkeeping and data products

A company keeps its books in its own AIMEAT: invoices, vouchers, VAT and fiscal years that lock. The company also gets a web address of its own, and it can publish tables that other programs, Excel and Power BI read straight from its AIMEAT.

#### Invoices and books: from the first invoice to the VAT report

**What you get:** the company's invoices and books in the same AIMEAT where the company already works, and an accountant who reads them without anyone sending files by hand.

**Ask your AI:** Ask your AI: “Build a twelve-month budget for my company in BUDJETTI and tell me in which month the cash runs out.”

#### A company address that points where you choose

**What you get:** the company has a web address of its own on your AIMEAT, `name.co.` followed by your AIMEAT's domain (on aimeat.io, for example `acme.co.aimeat.io`). You decide what the address shows and who may open it.

**Ask your AI:** Ask your AI: “Register my company, write a one-page company site, publish it at the company address and let anyone open it.”

#### Data products: tables with versions, a product sheet and a live feed

**What you get:** a table your company keeps, such as a price list or the results of a survey, published so that a program, an AI or a spreadsheet reads it straight from your AIMEAT, in the version it asks for.

**Ask your AI:** Ask your AI: “Publish our price list as a data package with a type for each column, and give me the address Excel connects to.”

### L15: Running your own AIMEAT

For whoever runs an AIMEAT for themselves or for a company, or bought one from the store: how to stand it up and back it up, the admin pages and settings, the look of its pages, the security and compliance pages, the classification a regulated organisation needs, and the package repository that brings software and its updates.

#### Stand up your own AIMEAT and keep it backed up

**What you get:** an AIMEAT of your own, on a laptop, on your company's server or as an instance bought from the store, and the knowledge of which version it runs and how to bring it back after a failure.

**Ask your AI:** Ask your AI: “Check which version my AIMEAT runs, whether a newer one is out, and what my backup leaves out.”

#### The admin pages: settings, people, schedules and morsels

**What you get:** one control panel for the whole installation at `/v1/admin` on your AIMEAT's address, which only the operator sees. Its menu groups the pages: the installation itself, Identity, Data, Infrastructure, Services, Integrations, Federation and Design.

**Ask your AI:** Ask your AI: “Read my AIMEAT's configuration and AI usage, and tell me which settings and costs I should look at first.”

#### The look: Themes & Styles, fonts and the front page

**What you get:** your AIMEAT's own pages (the home, the chat, Settings & Controls and the admin pages) in your company's colours and typefaces, and a front page that shows what you want visitors to see. Published apps and the public pages keep their own look.

**Ask your AI:** Ask your AI: “Copy the AIMEAT theme, make a style in our company colours and fonts, and show it to me before you offer it in the look picker.”

#### Safe and lawful: security, privacy and the compliance register

**What you get:** one place that says what was refused, what waits for your decision and what the law asks of you, and controls whose use the people concerned can see.

**Ask your AI:** Ask your AI: “Read my AIMEAT's security overview and tell me which incidents wait for my decision.”

#### For a regulated organisation: security classes, local models and regulated data

**What you get:** classification that follows the rules a public body, a hospital or a defence supplier works under, and content that no outside AI ever reads. The basics (levels, who sees what, exceptions) are in [Classify what is sensitive, and decide which AI reads it](#les-l9-classification); this lesson is what you add on top as the operator.

**Ask your AI:** Ask your AI: “Show me which security classes my AIMEAT offers, which are switched on, and which content only a local model may read.”

#### Packages and the store: sell software to other AIMEATs and keep it updated

**What you get:** software that you or a partner made, installed on a customer's AIMEAT in one step and kept updated from yours, with the money going through an ordinary checkout. How a package installs is in [Packages: install ready systems, run instances, build your own](#les-l4-packages); this lesson is the side of whoever sells it and runs the repository.

**Ask your AI:** Ask your AI: “List the packages my package repository offers this AIMEAT, and which installed packages have an update waiting.”

#### Sell your apps as a set

**What you get:** apps you have already published become one product that a customer buys and installs in one go: the apps, the shared place and the workspaces they keep their records in, the agents they bring and your default settings.

**Ask your AI:** Ask your AI: “Make a dry run of a set from my apps shop.html and backoffice.html, and tell me what a buyer will be asked and which problems remain.”

### B0: The value stack: where you are, where you're going

A single picture of how AIMEAT is built and how an organisation grows on it. You read it from the bottom up: from the first contact to working with others without depending on anyone. The business layers that follow build on this map.

#### Start

You arrive with the AI you already use: ChatGPT, Claude, Gemini or another. AIMEAT connects to it, and you keep the tool you know. Your AI chat works in AIMEAT directly over MCP, the connection that lets an AI use your AIMEAT. When your AI cannot connect, you use a ready-made app or copy a ready-made prompt into your own chat. You start where you already are.

#### Set up and organize

You give your work a structure. Organisms, the shared places where people and their AIs work together, are the top level. Workspaces sit inside them, and records, documents and workspace apps sit inside the workspaces. People and AI agents join as members, each with the access they were given. This is the emergent stage: things start to take shape.

#### Cowork, evolve & automate

People and AI agents work side by side in the same data, and the data is yours. Routine work starts to run on its own, raw information is refined into something useful, and decisions are made on it. This is the evolve stage: the structure becomes the everyday way of working.

#### Data sovereignty & value

Now the value shows: results delivered, work saved. You can follow what happened. The usage ledger records what each agent's AI calls cost, by day, agent and model, and content an AI made says how it was made. AI use is in view instead of hidden, and the data stays yours throughout. This is the transform stage, and here the customer value ladder begins.

#### Goal: strategy and leading

Ownership brings control and measurement. You can lead by a strategy because the data and the AI are yours and run where you choose. Governance follows from owning what you run, so you do not buy it as a separate product and attach it afterwards.

#### Result: connected, not dependent

The outcome at the top: you work with others and stay independent. You are connected to partners, customers and their AI, and information and work move between your organisations as you allow. No single tool or provider holds you, and every party keeps its own data in its own AIMEAT.

### F: Servers and federation: from one machine to a network

A node is one running AIMEAT, and the same node software runs everywhere: on a laptop, a department server or a company's central server. Federation connects nodes so that a sign-in can work across them when you and their operators allow it, while every node keeps its own data. Five diagrams, from a single node to a network between companies.

#### One node: the whole team on one machine

The simplest setup: one node with everyone on it. The node is the center: it holds the accounts, the data and the app catalog. **Users** attach to the node, each with an identity of their own (a GHII, in the form name@node). **Apps** are served by the node to the people who use them. **Agents** are authorized by users, so each agent belongs to a person, and they do their work through the same apps and data. One machine and clear ownership: every piece of data has an owner, and every agent has a person behind it. aimeat.io itself runs this way, and the same software runs on a laptop.

#### Department nodes: one company, federated

When a company grows, departments get their own nodes, each with the exact same structure as the single-node setup: its users, its apps, its agents, its data. **Federation** connects the nodes: every connection is between approved peers and every federated message is signed, so a node trusts only what it has chosen to trust. An employee's account lives on their department's node (their **home node**). What crosses between the nodes today is the public catalogue and the public records someone has let go to peers with a federation consent. Signing in on another department's node with the same account is possible, but it is off by default: the person allows that node to sign them in, the link between the two nodes carries sign-ins, and the other node's operator accepts sign-ins from its peers. Work requests do not cross nodes yet. Data does not pool anywhere on its own: each department keeps its own, and shares deliberately.

#### Personal nodes: private work, connected upward

The same node software runs on an employee's own computer. A **personal node** gives you the whole set (your user, which is you, your apps and your agents) for private work that stays on your computer. The AIMEAT desktop app can run one for you, together with the AI models that run on the same computer. A personal node anchors to the department's node through a tunnel, and that node keeps your messages and sends your push notifications while your computer is off. The department node can in turn peer with the company node: a chain of **local → department → company**, where each level owns its own data and passes on only the public records it has given federation consent.

#### Between companies: a partner network of nodes

The same pattern reaches past company borders. Each company runs its own node (or its own internal network of nodes) with its own users, applications and agents. **Federation between companies** connects them as peers: their public catalogues sync, public records with a federation consent are replicated, and a person may sign in on the partner's node when they, the link and the partner's operator all allow it (off by default). Work requests between nodes are not built yet, so to work on the same task today, a partner signs in to the node that holds the shared place, for example with their own company's Microsoft account when that node allows it. Every company's data stays on its own infrastructure, under its own ownership. This is where the value stack's crown points: their business and AI beside yours, as partners, connected but not dependent.

#### Sign-in and IAM: three layers of control

Identity has three layers, and they work together instead of competing.

### B1: Your company on AIMEAT

A business runs its own AIMEAT, bought from the store or installed on its own server, and can keep it for its own people and the partners it approves, who sign in with their company's Microsoft accounts. It takes payment through the checkout on its own Stripe account or by invoice, and sells agent offers, app tools and apps, each with an optional money price and a pacing amount.

#### Your company runs its own AIMEAT

**What you get:** your company's knowledge, agents, apps and sales in one AIMEAT that belongs to the company. No other company's account sits between you and your customers.

**Ask your AI:** Ask your AI: “Make a company address for [company name] and publish a front page that says what we do and how to reach us.”

#### Getting paid on your own account: card, invoice and the checkout

**What you get:** the money from a sale goes to your own account, and the buyer can be a person, your customer's AI or an outside shopping agent.

**Ask your AI:** Ask your AI: “Is my Stripe key stored and is invoice on? Give me the link to the Wallet page for what I still need to set.”

#### What a company can sell

**What you get:** your services, your app's tools and your apps reach paying customers, people and AIs alike, without a shop to build.

**Ask your AI:** Ask your AI: “List what I sell, with each item's money price or Free and its pacing amount, and suggest what to price next.”

#### Your company's own AIMEAT, signed in with your company accounts

**What you get:** one AIMEAT for your company. Your people sign in with the Microsoft work account they already have, the partners you approve sign in the same way, and nobody else sees anything.

**Ask your AI:** Ask your AI: “Make our AIMEAT a company AIMEAT where new accounts come only through Microsoft sign-in, content is for signed-in accounts only and classification is on for everyone, then show me which Microsoft directories may sign in.”

### B2: Making your APIs AI-native

Turn your own API, or your product's, into one that agents and AIs can find, use and pay for, billed per call or in other ways. A working example is PRH Yrityshaku, and the accelerator that builds it is TURBO.

#### An AI-native API: PRH Yrityshaku as the example

A regular API serves code. An **AI-native API** also serves agents and AI chats: it is discoverable, teachable and purchasable. The recipe has three parts:

**Ask your AI:** Ask your AI: “Look up [company name] with PRH Yrityshaku and give me the basics.”

#### TURBO: the accelerator that builds it

**TURBO** (turbo.apps.aimeat.io) is the accelerator that makes an API AI-native. It builds the recipe of the [previous lesson](#les-b2-api) around your API (the app, the skill that teaches agents to use it and the manifest of sellable tool calls), systematically and fast. PRH Yrityshaku was built with it.

**Ask your AI:** Ask your AI: “Compare our API with the PRH Yrityshaku example and list what an AI-native version of it would need.”

#### Billing for usage: the models

An AI-native API can bill through several models, and the payer can be a human, an agent or an AI. Each item has an optional money price (without one it is Free) and a pacing amount in morsels per use:

**Ask your AI:** Ask your AI: “Suggest a billing model for my API: which calls get a money price per call, which stay free, and what to sell as a derivative.”

### B3: Your own AIMEAT from the store, or a managed closed environment

Buy your own AIMEAT from the store and it is set up automatically once the payment is confirmed: your own database and storage, encrypted backups every night, updates done for you, and AI on the shared credit or your own key. When control is non-negotiable, a managed closed environment runs on a server reserved for you or on your company's own infrastructure.

#### Your own AIMEAT from the store, ready on day one

**What you get:** an AIMEAT of your own, running and looked after, without an infrastructure project. You pick a name and a size at the store (shop.apps.store.aimeat.io), pay, and the environment is set up automatically once the payment is confirmed.

**Ask your AI:** Ask your AI: “Make me an agent in my new AIMEAT that [job] every morning, and show it to me before it starts.”

#### A closed, managed environment

When a shared plan isn't enough, for data, regulatory or risk reasons, the alternative is **your own closed environment**: the whole system in your custody, inside your boundaries. In the store it is the **Managed** service, arranged with us before delivery.

**Ask your AI:** Ask your AI: “List what our company requires of a closed AIMEAT environment, for data, regulation and partners, so that I can bring the list to the planning conversation with AIMEAT.”

#### AI on your AIMEAT: the shared credit and a key of your own

**What you get:** AI that works from the first day on the AIMEAT you bought from the store, and a clear answer to who pays for each use. AI is paid either by your AIMEAT's shared AI credit or by a person's own key.

**Ask your AI:** Ask your AI: “Tell me who pays for AI use on my AIMEAT, and give me the link to the AI page where I add my own OpenRouter key.”

### B4: Agent-facing business

After portals: self-service and partner portals made AI-native, where the portal as such disappears. Agents carry the traffic between the parties, information stays current and answers come fast.

#### After portals: agent-facing self-service and partnerships

A traditional self-service or partner portal is a place where a person signs in to fill in forms. In **agent-facing business the portal as such disappears**: the service is still there, but both parties' AIs and agents use it. The person asks their own AI, and the work gets done.

**Ask your AI:** Ask your AI: “Go through our customer and partner processes and suggest which would move to the agent-facing model first.”

### B6: EXCHANGE: the marketplace where apps buy data

A marketplace where apps buy the data, services and agent work they need, by contract. Every call is counted, a budget cap protects the buyer, and the provider sees who uses their data. The money goes to the provider: a contract's calls are invoiced, a one-off purchase is paid by card on the provider's own Stripe account or by invoice, and the operator keeps a share. Morsels only pace the use.

#### Sell data, services and agent work

**What you get:** buyers for your data, your app's tools and your agent's work, with a contract that counts every call and pays you, without a shop of your own.

**Ask your AI:** Ask your AI: “I want to sell data or a service on EXCHANGE: build a provider extension or list my app's tool, and publish it as an offering with an input and output schema and a price.”

#### Buy for your app, or post a need someone fills

**What you get:** the data or service your app is missing, under a contract with a budget cap, without asking anyone for an API key.

**Ask your AI:** Ask your AI: “My app needs data X: run the whole chain on EXCHANGE, find an offering that fits or, when one is only close, contact the provider or build a converter, and open a contract with a budget cap.”

#### Money and trust: counting, payment and morsels

**What you get:** a market where real money moves and both sides see the same numbers. Its economy has three layers:
- **Counting, watched by both sides.** Every call draws on the buyer's budget. Under My contracts the buyer sees the spend, the calls and the operator's share; in the Provider console the provider sees who uses their data, how much, and what is owed to them. Both look at the same meter.
- **How the money moves.** When a contract has a money price, each call records what the buyer owes the provider, and the sum is invoiced. A one-off purchase at the checkout is paid by card on the provider's own Stripe account or by invoice. The operator keeps a small share of each counted call, at a rate the operator sets, and the provider keeps the rest. Stablecoin payment (x402, in USDC or EURC) is off by default, and on aimeat.io it runs on a test network whose tokens have no money value.
- **Morsels pace the use.** They are counted in whole numbers and are not money; the money moves its own way. Of the morsels a use takes, part burns, the operator keeps a share and the provider gets the rest.

**Ask your AI:** Ask your AI: “Show the spend and budgets of my EXCHANGE contracts, and, as a provider, what is owed to me and by whom.”

#### The same market inside a company: free, but visible

**What you get:** a map of which app uses which data inside your company, kept up to date by itself, also when no money moves.

**Ask your AI:** Ask your AI: “List our team's data source on EXCHANGE with no money price, then show me which apps use which internal source and how much.”

#### After the sale: payment, delivery and AI shoppers

**What you get:** once the buyer says yes, the money lands on your own account and the delivery follows in a set order. This is how.

**Ask your AI:** Ask your AI: “Check whether my own Stripe key is set, and tell me which of my offers and app tools AI shopping agents can buy, and why the rest are left out.”


> **AI provenance** — AI-assisted (claude-opus-5-5), human editorial review, 2026-11-07T14:11:00.374Z · record: https://aimeat.io/v1/provenance/c1db8461-5fba-45cf-8478-86b8ac83d78f


## Agent affordances

- Act — WebMCP tool listing for this app: https://aimeat.io/v1/apps/happydude500001/experience-center.html/webmcp
- Learn — skills bound to this app: https://aimeat.io/v1/apps/happydude500001/experience-center.html/skills
- App catalog on this node (this app: happydude500001/experience-center.html): https://aimeat.io/v1/appcat
- Register an agent on this node (RFC 8628 device flow): https://aimeat.io/auth.md