The configure-ai-agents command now copies extensibility artifacts (commands, skills, subagents) from the central nx-ai-agents-config repo for non-Claude agents. OpenCode is now supported with project-level MCP configuration, and Claude uses the marketplace plugin with MCP included. Configs are cached by commit hash for fast, reproducible installs with automatic cleanup of old versions.

Nx docs now support HTTP content negotiation for LLM and CLI tool access. Request any docs page with Accept: text/markdown header to receive raw markdown content directly from the canonical URL. No .md suffix required, no redirects. Browsers continue receiving HTML as expected.

Nx now includes a Task IO Service that tracks process IDs, hash inputs (files, env vars, runtime commands), and output files for each task. This gives you full visibility into what a task executed, what it consumed, and what it produced—making it easier to debug hung processes, audit cache behavior, and build custom tooling on top of Nx's execution model.

Nx documentation now supports LLM discovery out of the box. Every docs page includes HTTP Link headers pointing to its raw markdown source, plus links to llms.txt and the new llms-full.txt, which compiles all 500+ documentation pages into a single file for AI consumption. Build internal chatbots, power RAG pipelines, or enhance your coding assistants with accurate, structured Nx knowledge.

When Nx runs batch tasks (multiple projects with a single executor), the TUI now groups them under one collapsible entry. You see a single Running/Success/Failure status for the batch, and can expand to view individual task results and terminal output. This reduces visual noise during large runs and makes it faster to identify failures.