AI News Summary - September 29, 2026

Claude Code Launches Multi-Repository Context Feature

Anthropic released a major update to Claude Code that allows developers to simultaneously work across up to 10 linked repositories with shared context. The feature uses an improved token efficiency algorithm that reduces context window usage by 40% compared to loading files individually. Early adopters report significant productivity gains when refactoring code that spans microservices architectures.

MCP 2.1 Specification Adds Bidirectional Streaming

The Model Context Protocol consortium published version 2.1, introducing bidirectional streaming capabilities that enable real-time data synchronization between AI models and external tools. The update includes native support for vector database connections and a standardized authentication layer that works across major cloud providers. Over 200 MCP-compatible tools have already committed to supporting the new specification within 60 days.

Open-Source Agent Framework "Conductor" Reaches 1.0

The Conductor project, an Apache-licensed agent orchestration framework, officially hit version 1.0 after 18 months of development. It features built-in support for tool chaining, automated error recovery, and cost optimization across multiple LLM providers simultaneously. The framework has already been adopted by three Fortune 500 companies for production deployments.

GitHub Copilot Workspace Adds AI-Powered Code Review

GitHub announced Copilot Workspace now includes autonomous code review capabilities that can analyze pull requests for security vulnerabilities, performance issues, and architectural anti-patterns. The feature integrates directly with CI/CD pipelines and provides actionable suggestions with specific code fixes. Initial benchmarks show it catches 35% more issues than traditional static analysis tools.