AI News Summary - July 2, 2026

Claude Code Introduces Multi-Repository Context Window

Anthropic expanded Claude Code's context window to support simultaneous analysis across up to 50 linked repositories, enabling developers to refactor code across entire microservice architectures. The update includes automatic dependency mapping and cross-repo symbol resolution. Early adopters report 40% faster resolution of complex architectural changes compared to single-repo workflows.

MCP 2.1 Specification Adds Streaming Context Support

The Model Context Protocol consortium released version 2.1, introducing bidirectional streaming for real-time context updates between AI models and external data sources. The specification enables agents to maintain synchronized state with databases, APIs, and file systems without polling. Major MCP server implementations from Vercel, Supabase, and DataStax are already compatible with the new streaming protocol.

OpenHands Agent Framework Reaches 1.0 with Enterprise Features

The open-source OpenHands project achieved its 1.0 milestone, adding multi-agent orchestration, secure sandbox environments, and audit logging for enterprise deployments. The framework now supports coordinated task distribution across specialized coding, testing, and documentation agents. Over 200 companies have adopted OpenHands in production since the beta launch in Q1 2026.

Cursor AI Ships "Time Travel Debugging" for Python and TypeScript

Cursor's latest update introduces AI-powered time travel debugging that automatically generates reproducible test cases from production errors. The system analyzes stack traces, variable states, and execution history to create minimal reproduction environments. The feature currently supports Python and TypeScript with Java and Go support planned for Q3.