6 concepts, live from D1

Concepts

Real, well-established Anthropic Console org-admin concepts, grouped by category unless filtered.

all categories organization workspace api_key roles cross_reference

organization

concept	description
Organizations	The top-level account and billing entity in the Anthropic Console. An organization owns billing, usage limits, and the set of workspaces and members beneath it. Every API key, workspace, and member ultimately belongs to exactly one organization.

workspace

concept	description
Workspaces (Console)	A grouping mechanism within an organization for organizing API keys and usage — typically used to separate spend/usage tracking and access between teams, environments, or projects (e.g. "production" vs "staging", or one workspace per internal team). Workspaces are a Console-level administrative construct, not a place where conversations or agent context live.

api_key

concept	description
API key scoping to a workspace	An API key created in the Console can be scoped to a specific workspace, meaning usage and billing attribution for requests made with that key roll up under that workspace rather than the organization's default/unscoped pool. This lets an organization see and control usage per team or environment without splitting into separate organizations.

roles

concept	description
Member roles and permissions (org-level)	Members of an organization are assigned org-level role types such as admin and member (and typically a billing-related role), which govern administrative capabilities like managing API keys, workspaces, and billing settings within the Console. Exact role names and permission granularity are phrased generally here rather than asserted precisely, since they were not independently re-verified in this pass.

concept

description

Member roles and permissions (org-level)

Members of an organization are assigned org-level role types such as admin and member (and typically a billing-related role), which govern administrative capabilities like managing API keys, workspaces, and billing settings within the Console. Exact role names and permission granularity are phrased generally here rather than asserted precisely, since they were not independently re-verified in this pass.

cross_reference

concept	description
How org/workspace structure relates to agent identity and permission policies	Org and workspace structure (this site) governs human administrative access to Console resources — who can create keys, see billing, manage workspaces. It is a separate layer from agent IDENTITY/attribution (which surface an agent acts under, service-account vs. personal-account — see subagentidentities.com) and from agent TOOLSET permission policies (which tools an agent may call and under what default posture — see subagentpermissions.com). An org can contain many workspaces, each of which may host agents running under distinct identities and permission policies.
Workspaces vs. Claude.ai Projects	A common point of confusion: a Console "workspace" and a Claude.ai "Project" are not the same thing. A workspace is an org-admin construct for grouping API keys and usage/billing inside the Anthropic Console (the developer/API surface). A Project in Claude.ai is a user-facing container for conversations, files, and custom instructions inside the consumer/chat product. They live in different products, serve different purposes, and having one does not imply or create the other.

concept

description

How org/workspace structure relates to agent identity and permission policies

Org and workspace structure (this site) governs human administrative access to Console resources — who can create keys, see billing, manage workspaces. It is a separate layer from agent IDENTITY/attribution (which surface an agent acts under, service-account vs. personal-account — see subagentidentities.com) and from agent TOOLSET permission policies (which tools an agent may call and under what default posture — see subagentpermissions.com). An org can contain many workspaces, each of which may host agents running under distinct identities and permission policies.

Workspaces vs. Claude.ai Projects

A common point of confusion: a Console "workspace" and a Claude.ai "Project" are not the same thing. A workspace is an org-admin construct for grouping API keys and usage/billing inside the Anthropic Console (the developer/API surface). A Project in Claude.ai is a user-facing container for conversations, files, and custom instructions inside the consumer/chat product. They live in different products, serve different purposes, and having one does not imply or create the other.