AI News Summary - May 12, 2026

Claude Code Launches Multi-Repository Analysis Feature

Anthropic has released a major update to Claude Code that enables simultaneous analysis across up to 10 linked repositories. The feature uses an improved context window architecture to maintain coherent understanding of complex microservice architectures and monorepo structures. Early adopters report 40% faster debugging times when working with distributed codebases.

MCP Connector Hub Reaches 500 Official Integrations

The Model Context Protocol ecosystem has surpassed 500 verified connector integrations, with new additions including SAP systems, Snowflake data warehouses, and Figma design files. The MCP Foundation announced that over 2,000 enterprise deployments are now using the protocol to connect AI agents to proprietary data sources. A new security certification program ensures all connectors meet enterprise compliance standards.

Open-Source AgentKit 3.0 Introduces Persistent Memory Layer

The popular AgentKit framework has released version 3.0 with a built-in persistent memory system that allows agents to maintain context across sessions without custom database implementations. The update includes vector storage optimization reducing memory overhead by 60% and new APIs for fine-grained control over what agents remember. Over 15,000 GitHub stars within the first 24 hours signal strong developer interest.

GitHub Copilot Workspace Adds Natural Language Refactoring

GitHub announced Copilot Workspace can now handle complex refactoring tasks described in plain English, such as "convert this REST API to GraphQL" or "migrate authentication from JWT to OAuth2." The system generates a step-by-step execution plan for review before making changes across multiple files. The feature is currently in limited beta for GitHub Enterprise customers.