Anthropic updates Claude with memory for all users, skills via LLM gateway, and Cowork task scheduling.
Rolling out persistent memory shifts Claude from a stateless query engine to a continuous context assistant, drastically reducing prompt friction for complex workflows. The addition of skills via an LLM gateway and scheduled tasks indicates Anthropic is aggressively building out enterprise-grade orchestration and agentic tool-use capabilities.
What happened
Anthropic has rolled out a major feature update to Claude, introducing persistent memory across chat histories for all users, including those on the free tier. According to the latest release notes, the update also includes new support for "skills" routed through an LLM gateway, as well as the ability to create and schedule recurring and on-demand tasks within Claude's "Cowork" environment.Technical details
The introduction of statefulness via chat history memory is a significant architectural shift. Claude can now automatically retrieve and build upon previous interactions, likely utilizing a vector database backend and RAG (Retrieval-Augmented Generation) pipeline for semantic search across user history.Additionally, the explicit mention of "skills via an LLM gateway" points to a formalized, modular tool-use architecture. Rather than relying solely on monolithic prompting for external actions, Claude is leveraging a gateway pattern. This allows the core model to route requests to specific APIs or functions (skills) while the gateway handles authentication, logging, and rate limiting. The addition of scheduled tasks in Cowork introduces a cron-like orchestration layer, allowing the model to execute asynchronous, time-delayed operations.