Anthropic

The Corma and Anthropic integration governs the Claude Console, where your teams manage Claude API workloads: members, workspaces, API keys and spend. Anthropic separates claude.ai organizations, where users chat, from Claude Console organizations, where developers build with the API, and since January 12, 2026, console.anthropic.com redirects to platform.claude.com. Seats in the Claude apps are covered on our Claude access management page; this page is about the API side.

Key takeaways

• Claude API usage is paid with prepaid usage credits, or monthly in arrears for some organizations, and priced per million tokens, so Console governance is mostly about keys and spend.
• Only organization Admins can invite and remove Claude Console members, and organization admins automatically get Workspace Admin in every workspace.
• When a user is removed from a Console organization, their personal API keys stop working, while service account keys and legacy workspace keys keep working.
• Only organization admins can create Admin API keys, and these keys carry full access to every endpoint that accepts them.

Which Claude Console keys survive a departure?

A new Console API key is linked either to its creator, as a personal key, or to a service account shared across users, and can be scoped to one workspace. Anthropic's Admin API guide spells out what happens at offboarding: personal keys stop working when their user is removed, service account keys keep working even if their creator leaves, and legacy workspace keys keep working too. A legacy workspace key belongs to no one, so it runs until it expires, is disabled or deleted, or its workspace is archived.

Spend follows the same structure. Workspaces can have spend limits set below the organization's limits, and usage and cost can be tracked per workspace. Admin API keys, which manage members, workspaces, invites and keys, have no selectable scopes, which makes them the most sensitive credentials in the organization.

How Corma governs Anthropic

• Members and roles under review: Console members and their roles, from User and Developer to Billing and Admin, go into automated access reviews.
• Offboarding that protects integrations: a leaver workflow in the IT Workflow Builder asks an organization admin to move integrations from the person's personal keys to a service account before the member is removed.
• Keys with an owner: service accounts and legacy workspace keys are non-human identities that outlive people, so each one needs a named owner and a regular check.
• AI spend in view: the Anthropic bill sits next to your other AI tools in Corma's AI spend management, with reporting, budgeting and forecasting.

Frequently asked questions

Does the Claude Console support SSO and SCIM?

Anthropic says single sign-on is available to Console organizations, with SAML guides for Okta, Entra ID, Google, OneLogin, JumpCloud and Duo, and its support articles indicate that Console SSO is enabled with Anthropic's Sales team. SCIM provisioning is available to Console organizations, except those joined to a Team plan's parent organization. See how SCIM and SAML differ.

How does Corma connect to Anthropic?

Through Anthropic's Admin API where it fits, or through Corma's browser-based agents in the Console. The Admin API cannot create keys or remove members with the admin role, so those steps stay in the Console. Corma governs other AI platforms the same way, such as OpenAI Platform.

Where is Corma data hosted?

Corma runs on AWS, with web servers and databases in data centers inside the European Union. Data is encrypted in transit and at rest, and Corma is ISO/IEC 27001:2022 certified.

Want to know which Claude API keys still run without an owner? Book a Corma demo.