anthropic-sdk-python v2.11.0
- Source
- github.com
- Author
- anthropic-sdks-releaser[bot]Top Viber
- Date

Context
Version 2.11.0 of Anthropic's Python SDK adds the methods a Claude Enterprise organization needs to manage Claude Code's company-wide configuration from its own scripts (release notes). Claude Code is Anthropic's coding , and managed settings are the configuration an organization pushes to the people who use it, such as permission rules or the models they may pick. According to Claude Code's documentation, these settings occupy the highest tier of its settings hierarchy, so a key set there overrides the same key in a developer's own settings files. An Owner edits them in the claude.ai admin console, and Claude Code fetches them at startup and every hour.
The new methods, added under `client.beta.organization.claude_code.managed_settings` in this commit, cover the group-based setup that the documentation describes for Enterprise plans. An organization can store a separate settings document for each RBAC group, a group of members used for role-based access control, rank those groups, and set default settings for members who belong to none of them. Each document replaces the stored one whole, and a member of several groups receives only the settings of the highest-ranked group. Every call needs an organization-level Admin API key with a read or write scope for Claude Code managed settings, plus a beta header, and the docstrings say the API is in beta and open only to Claude Enterprise organizations.
The step most likely to catch an administrator out is switching the organization from one organization-wide configuration to group mode. The configuration method's docstring says that after the switch, a member in no listed group receives the default settings, or none if no default exists, instead of the organization settings. If no group has settings yet, the switch copies the organization settings into a new default so members keep what they had. Once any group has settings, the group method's docstring says the switch no longer makes that copy. A script that writes group settings first and then flips the mode can therefore leave every developer outside those groups with no server-managed policy unless it sets the default first, although settings deployed to the device itself still apply, according to Claude Code's documentation.
The release's one bug fix concerns requests that the SDK signs with AWS's SigV4 scheme. The SDK used to send a space in a query parameter as a plus sign, and the fix sends it as `%20` so the signature matches the query string. The remaining entries add an error type to the Admin API's plugin methods, change how structured-output calls build their schema for performance, and edit documentation.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
- sandbox — An isolated environment where AI-generated code or agent actions run without being able to touch anything real.
- structured output — Forcing a model's response to match a schema, so downstream code can parse it instead of guessing at prose.
Checking sign-in…
Loading comments…





