AI News Summary - April 2, 2026

GitHub Copilot Workspace Adds Multi-Repository Context

GitHub announced Copilot Workspace now supports analyzing and editing across up to 10 linked repositories simultaneously. The update leverages an enhanced context window that maintains semantic relationships between codebases, making it particularly useful for microservices architectures. Early access users report 40% faster cross-repo refactoring tasks.

MCP 2.1 Specification Brings Streaming Data Sources

The Model Context Protocol consortium released version 2.1, introducing native support for streaming data sources and real-time context updates. This enables AI assistants to maintain live connections to databases, APIs, and file systems without repeated polling. Major implementations from Anthropic, Microsoft, and Google are expected within the next two weeks.

Open-Source Agent Framework "Conductor" Reaches 1.0

Conductor, a TypeScript-first agent orchestration framework, officially launched version 1.0 after 8 months in beta. The framework provides built-in MCP support, deterministic execution modes, and a visual debugging interface for multi-agent workflows. The project has already gained 15K GitHub stars and backing from Sequoia Capital.

Claude Code Now Supports Test Generation from Natural Language

Anthropic updated Claude Code with an experimental feature that generates comprehensive test suites from plain English descriptions of expected behavior. The system analyzes existing code patterns in your repository to match your testing style and framework preferences. Initial benchmarks show 85% accuracy on standard unit test scenarios.