← All topics

Field 01 / Foundations

Define the system.

Before choosing a model, decide what the service is allowed to know, say, and change. These guides separate product scope from interface polish and turn a vague assistant idea into an accountable service.

The question to keep asking

Can someone explain the service without naming its model?

A useful review starts here

  • One user job has a clear completion state.
  • Every consequential decision has an accountable owner.
  • The person can see what the assistant cannot do.
Editorial prompts, not certification or a complete risk assessment.
  1. 01

    Name every component ↗

    Map the channel, identity, state, evidence, tools and human owner.

  2. 02

    Write the promise ↗

    Make supported jobs, limits and alternative routes visible.

  3. 03

    Protect the identity boundary ↗

    Move verification into an appropriate secure surface.

See the source / related interface context

A shared connector is not a trust policy.

Interoperability and authorization are different layers.

Open the case file ↗This source capture illustrates a related interface—not evidence that the vendor passed these review checks.
Anthropic’s orange-and-white illustration of a connection between an assortment of shapes and a single shape
Official announcement illustration · Anthropic · Original source ↗Local visual review · not cleared for production

The complete foundations desk

Go deeper.

Search this topic ↗

Keep the system connected

Conversation design Knowledge & actions Put the reading to work

Find your next good decision.

Start typing to explore the guides.

76 sourced guides · Escape to close