Vercel CLI shipped a major redesign of blob store management, graduating metrics to general availability, and expanded experimental services support. The vercel blob command now requires explicit --access flags, replaced deprecated subcommands with create-store, delete-store, and list-stores, and added empty-store with confirmation flows that surface connected projects. The vercel metrics CLI shed its beta label while simplifying metric selection and cleaning up schema output. Behind the scenes, experimental services gained support for dynamically specifying crons from Python, configurable environment prefixes per service, multiple topics for workers, and auto-detection from Railway configs. New CLI commands arrived for managing deploy hooks, edge configs, project checks, speed insights, web analytics, tokens, team SSO, OAuth apps, and project-level routing rules with AI-powered generation. The CLI also strengthened observability with deployment checks support in deploy --prod, showed check run logs inline with deep-links, and cached authenticated user IDs to reduce API requests.
March shipped agent-focused enhancements and expanded routing capabilities. The CLI auto-installs agent tooling (Claude, Cursor) and accepts structured JSON prompts for plugin approval via vercel login, vercel link, and vercel deploy. New vercel routes command lets developers manage project-level routing rules—rewrites, redirects, response headers—with optional AI generation. Build-utils exposed hash and lambda utilities for shared consumption, while Python gained Django static file handling and Celery worker declarations with broker_url="vercel://". Services configuration matured with multiple topics per worker, env prefix support, and subdomain routing.