Anthropic Introduces New Features and Defaults in OpenClaw Integration
Anthropic has announced several updates to its integration with OpenClaw, including new features and default settings for its API keys and Claude CLI usage. The company's models now default to adaptive thinking in OpenClaw when no explicit thinking level is set, but users can override this setting per-message or in model parameters.
OpenClaw also supports Anthropic's prompt caching feature, which allows for faster response times and improved performance. By default, the cache retention period is set to 5 minutes for all Anthropic models using API key authentication, but users can override this setting by configuring a different cache retention period in their model settings.
Anthropic has also made its 1M context window beta available in OpenClaw, allowing users to access supported Opus and Sonnet models with extended context windows. However, this feature is still in beta and requires explicit configuration per model. Additionally, Anthropic's Claude CLI backend is now supported in OpenClaw, but users are advised to use API keys for production environments due to potential changes in Anthropic's policies.
To troubleshoot issues related to 401 errors or token invalidation, users can run the openclaw models status command to verify their active authentication profile and re-run onboarding or configure an API key if necessary.
