Claude Console adds an IP allowlist

On 5 November 2026 Anthropic announced, in beta, an organisation-level IP allowlist for the Claude Console. Requests from addresses that are not on the list are refused. The aim is to limit where a leaked API key can be used.

What does the IP allowlist do?

According to the platform release notes, an organisation can add IPv4 and IPv6 addresses or CIDR ranges, up to 1,000 entries per organisation. Requests to covered services from addresses not on the list are refused. The feature is in beta. Which services are covered and how to set it up are explained in a separate document the note points to.

Why does it matter?

The most common accident in small teams that use AI services is an API key ending up in the wrong place along with the code. If a key leaks, an attacker can use it from their own machine and you pay the bill. An IP restriction acts as a second lock in that scenario: even if the key is stolen, only requests from your own servers get through.

What to watch for

A restriction like this works comfortably only where outbound addresses are fixed. On serverless infrastructure outbound addresses can change, so building the list correctly can be hard; know which addresses your application uses before turning the feature on. Keeping keys out of code and rotating them regularly remain separate safeguards.

If you are thinking of connecting AI services to your business, see how we work on our custom projects and automation page.