Claude's Redesign: Parallel Code Panels & Cloud Routines
Anthropic's Claude desktop now supports up to 4 parallel Claude Code panels with per-panel terminals and web previews, plus cloud routines for scheduled tasks via cron or API triggers—no local machine needed.
Parallel Panels Supercharge Multi-Project Coding
Run up to four Claude Code instances simultaneously in split-view panels within the same desktop app, mimicking terminal multiplexing but with integrated previews and controls. Each panel gets its own dedicated terminal, real-time task updates, and web app previews—eliminating browser switches for seamless dev workflows. Pin key chats for quick access, switch tasks across panels (they run independently in the background), review diffs inline, and process PRs directly without leaving the interface. This kills context-switching costs: edit CSS for white text on an 'index all documents' button in one panel while redesigning a UI based on Linear's template in another. Trade-offs include occasional slowdowns on API calls and UI bugs like non-resizable dividers, but it positions Claude as an IDE killer for solo devs handling multiple repos.
Cloud Routines Automate Repetitive Tasks
Shift scheduled tasks (formerly local-only) to Anthropic's infrastructure as 'routines' that run hourly, daily, or via custom cron jobs regardless of your machine's state. Define a routine with a name, prompt (e.g., 'categorize inbox and draft responses'), repo/folder, model, connectors (like email or news feeds), and triggers—including API endpoints for on-demand invocation from other apps. Templates accelerate setup: email triage prioritizes inboxes with drafts for urgent items. Use cases include daily video idea generation from aggregated news or email reports. Leverage installed plugins for external integrations, ensuring routines execute reliably without local uptime. This enables always-on automation, like content pipelines, but ties you to Anthropic's compute and potential costs.
Pricing Pivot Warns of Token-Based Billing
Anthropic now bills enterprise firms by AI usage tokens amid compute shortages, ditching flat subscriptions—individual tiers may follow. Echoes OpenAI's Codex restrictions: plus subscribers hit 5-hour waits after 2-3 prompts despite unchanged weekly limits, rendering it demo-only. Expect similar caps on Claude's max plans, prioritizing heavy users for pay-per-use while light ones face throttling. Builders should monitor for impacts on routine-heavy workflows, as cloud execution amplifies token burn.