Skip to main content
Post

You can now set up an IP allowlist for your organization in the Claude Console, in beta

Anthropic added IP allowlisting in beta for organizations in the Claude Console. Requests to covered services from addresses that aren't on your list are refused, so a leaked API key only works from networks you've approved.

An organization admin or owner adds IPv4 or IPv6 addresses, or CIDR ranges, on the Identity and access page. The list holds up to 1,000 entries and covers the whole organization. Ranges can be no broader than /8 for IPv4 or /32 for IPv6, and they need to be public addresses, not private ranges like 10.0.0.0/8. Changes can take up to 15 minutes to apply.

The allowlist covers Console access and the Messages, Message Batches, Token Counting, Models, Files, and Skills endpoints. It also covers the Managed Agents endpoints (Agents, Sessions, Environments, Vaults, Memory stores), plus the Admin API and Compliance API. A blocked API request gets a 403 permission_error, and blocked people see a page saying their network can't access the organization.

The docs recommend ticking the checkbox that lets admins reach the settings page from any IP, so a bad list doesn't lock everyone out. The tradeoff is that someone who compromises an admin session could edit the list remotely. The page checks your current address when you save, but it can't check your servers, CI, or teammates, so list those networks first and watch for 403s from jobs that only run daily or weekly.

It doesn't apply to Claude in Amazon Bedrock, Claude Platform on AWS, Claude Models on Google Cloud, or Claude in Microsoft Foundry. Anthropic also says an allowlist limits where a key works but not for how long, so it suggests short key expirations or Workload Identity Federation alongside it. Enterprise organizations have a separate allowlist for the Claude apps.

Loading replies…