Anthropic Introduces Rate Limits API: Administrators Can Now Programmatically Retrieve Rate-Limit Configuration for Their Organization and Workspaces

Anthropic announced on April 24, 2026 in its release notes a new Rate Limits API, part of the broader Admin API, which allows administrators to programmatically retrieve the rate-limit configuration set for their organization and individual workspaces. This is a natural extension of functionality that previously existed only within the Claude Console graphical interface.

Which Endpoints Does the New API Expose and What Do They Return?

The API exposes two primary endpoints, both accessible through https://api.anthropic.com. GET /v1/organizations/rate_limits returns all rate limits applied at the organization level for the Messages API and associated resources (limits for Claude Managed Agents are not included). GET /v1/organizations/workspaces/{workspace_id}/rate_limits returns only the overrides set for an individual workspace — anything absent from the response means the workspace inherits the organization-level value.

Both endpoints support an optional group_type query parameter for filtering by category. Valid values are model_group, batch, token_count, files, skills, and web_search. The organization endpoint additionally accepts a model parameter that returns the group to which a given model belongs.

Each record represents a single rate limit group. For groups of type model_group, the models field lists all model IDs and aliases sharing the same quota (e.g., Opus 4.5, 4.6, and 4.7 share the same RPM quota). The limits field contains {type, value} pairs where type can be requests_per_minute, input_tokens_per_minute, or output_tokens_per_minute. The workspace endpoint additionally attaches an org_limit field to each limit, making it easier to compare and audit overrides.

Three Practical Use Cases

The endpoints require a dedicated Admin API key with the prefix sk-ant-admin, separate from a standard API key. Only organization members with the admin role can generate one through the Claude Console interface, meaning access to limits does not need to be embedded in runtime code — it can be kept isolated in a CI/CD or monitoring process.

Anthropic explicitly highlights three scenarios in the documentation. The first is gateway and proxy synchronization — a proprietary API gateway reads the current limits on startup and on a schedule, rather than hardcoding values that drift when Anthropic changes them. The second is internal alerting — comparing data from the Usage and Cost API against configured limits for early warnings before 429 errors start being returned. The third is configuration auditing — verifying that workspace overrides still match what the provisioning automation expects.

For teams running Claude integrations for multiple clients through the same workspaces, the third point is particularly relevant — it is now possible to have a Terraform or Ansible job periodically verify that no one has manually changed a limit outside the agreed configuration. The API does not support write operations; workspace limits still need to be changed through the Console.