`ant apply` is now available in the ant CLI, letting developers declare Claude-managed agent environments, agents, skills, memory stores, and deployments as repository files and keep them in sync with the API.
AGENTS.md support is confirmed as a planned addition to Claude Code, in response to a user request.
The Claude Code team is exploring Function Hooks, a new system for extending and customizing Claude Code behavior. The feature hasn't shipped yet; feedback is being collected on a GitHub issue.
Weekly usage limits have been reset for all Claude Max plan subscribers, timed to coincide with the Fable 5.1 release and a long weekend.
Retailers using shopping agents built on Claude are reporting carts up to 35% larger and checkout rates 60% higher. Anthropic shipped a blueprint covering a shopping agent, a merchant agent, and production-tested guardrails.
Direct links to the shopping agent blueprint and accompanying blog post for building commerce agents on Claude.
An Anthropic team member follows up with a developer already using Claude to trace client journeys and trigger emails and SMS, inviting them to check out the newly shipped commerce agent blueprint.
Weekly usage limits have been reset for all Claude Max plan subscribers, timed to coincide with the Fable 5.1 release and a long weekend.
Claude Code on desktop includes a built-in browser that lets it edit code, open a live preview, click through the app flow, verify changes, and self-correct bugs, all in one loop without switching tools.
A developer whose Claude Code OSS license ended thanked Anthropic for enabling projects like a Duke Nukem port and Quartz .NET 4.0. Lydia Hallie confirmed that reapplying is an option.
A developer who was invited to the Claude Code OSS program but cannot access it reached out publicly. Lydia Hallie offered to look into it directly if the user DMs their email.
The recommended first-pass approach for debugging failing hooks is simply to ask Claude "why did the hook fail?" rather than digging through logs manually.
Fable 5.1 access has expanded; it is now available not just via API, confirmed as of yesterday.
The fix for effort changes breaking prompt cache applies only to Fable 5.1; Opus is unaffected by the improvement.
Fable 5.1 is accessible in both the app and CLI; users should ensure they are on the latest version to get app access.
Fable 5.1 access has been extended to subscribers as of yesterday, not just API users.
Switching the /effort level mid-session on Fable 5.1 no longer invalidates the prompt cache, fixing a longstanding friction point for users who adjust effort dynamically.
Changing the effort level mid-session on Fable 5.1 no longer invalidates the prompt cache, a fix that landed yesterday.
The recommended effort setting for Opus 5 is high or extra-high, in contrast to Fable 5.1 where medium is now the suggested default.
As of v2.1.251, running /effort saves the level for the active model rather than globally. Settings can also be configured directly via modelSettings in settings.json, so each model remembers its own preferred effort level.
