Get started

How Archetype works

The ideas behind Archetype: what a brand is, what it holds, how it reaches your agents, and how it stays current.

Archetype holds one source of truth for your brand and serves it to the AI agents your team builds with. This page explains the model in one pass, with links to the pages that go deeper.

A brand is a workspace

Everything in Archetype belongs to a brand. A brand has its own library, its own team with roles, its own plan and its own address for agents.

You can hold several, for example one per product or per client. Switch between them from the brand menu in the top bar, and add another from Settings → General with New brand. Each brand has its own plan, and the free trial is once per person (see Plans and trial).

People join a brand by invitation from Settings → Team, as an Owner, Admin, Editor, Creator or Viewer. Connected agents don't take a seat. See Members and roles.

What a brand holds

The sidebar lists the brand's areas under Brand, from its foundations out to the things built on them. Icons has no row of its own: it is a card on Foundations.

AreaWhat it holds
FoundationsColors, gradients, typography, spacing, radius, shadow and logos, served as design tokens, plus the Icons card.
ComponentsYour UI kit, as real HTML and CSS bound to your tokens, with properties and examples.
IconsThe icon sets to use, so agents never reach for arbitrary icons. Opened from Foundations.
AssetsPhotos, illustrations, files and PDFs, described and tagged so agents can find them.
ArtifactsBespoke brand code: backgrounds, patterns, generative graphics and motion.
TemplatesFinished layouts for social, web, app, email and documents, with slots to fill.
KnowledgeBRAND.md, company facts and voice documents: the written half of the brand.

Each item carries a status: Draft, In progress, In review, Complete or Deprecated. Only finished work leaves the app. An item marked Complete, or one nobody has given a status, reaches agents; everything else stays inside Archetype until it is ready. See Item states.

How it reaches agents

One MCP server per brand

Each brand has its own MCP address, https://www.heyarchetype.com/api/mcp/<brand-slug>. You add it to an agent once and sign in in the browser, where you pick the brand and choose whether the agent may change it. Someone with several brands adds one connector per brand. See Connect an agent.

The server gives the agent a small set of tools: get_brand to read the brand, search_brand to find anything in it by keyword, check_against_brand to grade work against it, generate_media to make new images and video for its library, create_campaign to have a whole campaign planned and made, and update_brand to change it when you ask. See MCP tools.

BRAND.md routes the work

get_brand with no arguments returns the brief, which is where every agent starts. It opens with BRAND.md, the brand's own routing: what the company is, the rules every task follows, and what each kind of work needs. You write BRAND.md in Knowledge under Start here.

Read an index, then the detail

get_brand reads one domain at a time. Called bare, a domain returns an index: every item with the handle that fetches it. Called again with a handle, it returns the item itself, such as a component's real HTML and CSS or an artifact's source to use verbatim. The domains are brief, info, tokens, assets, components, icons, artifacts, templates, files, voice and schema.

When the brand has no value for something the agent needs, the server tells it to use the nearest value the brand does define and to say so in its output. A gap gets reported instead of filled in quietly.

The same brand also leaves Archetype as files, for repositories and tools that don't connect over MCP: a brand markdown file, tokens as JSON, a downloadable bundle and installable skills. See DESIGN.md, BRAND.md and skills and The CLI.

The loop

A brand in Archetype is never finished. The work moves in a loop:

  1. Import. Bring the brand in from a website, a Figma file or your code, or let a connected agent import it. See Import overview.
  2. Review. Look over what came in, fix what the import got wrong, and fill the gaps.
  3. Set states. Mark what is ready as Complete, and keep work in progress out of your agents' reach.
  4. Agents use it. Every connected agent reads the current brand on every call. Change a token in Archetype and the next build uses it.
  5. Check. Agents grade their own work with check_against_brand, and you can hold a page, file or screenshot up against the brand on the Check page. See Check.
  6. Agents propose changes. An agent you allowed to edit can add to the brand when you ask. By default each change waits as a draft in History until someone applies or discards it. See Changes from agents.
  7. History keeps it all. Every edit, by a person or an agent, is logged with its before and after, and can be undone. Save named versions of the whole brand to compare and restore. See History and versions.

You can also make things inside Archetype with the same brand: designs, images and video on the Create page, or a whole campaign of them from one brief. See Designs and Campaigns.

Next steps

Something missing or wrong on this page? Write to support@heyarchetype.com.