Anthropic Is Testing Hub Mode, a Control Center for Claude's AI Sub-Agents

Anthropic is quietly testing a feature called Hub Mode in Claude's mobile app, an internal option marked with a planet icon that appears designed as a centralized control center for managing AI sub-agents. The discovery was made by TestingCatalog, which spotted the hidden toggle in the app. The feature is flagged as internal, and enabling it produces no visible changes — the underlying functionality is not operational yet. A description tells users the mode will apply to new tasks and requires an app restart after activation, standard language for features still in early development.

More telling evidence comes from Claude Code on the web. Navigating to a Hub destination reveals an interface listing sub-agents; selecting one opens a page stating that Claude will generate a task board for the user and their agents. That single detail — a task board coordinating multiple agents — signals a shift away from Claude's current single conversational thread toward a future where users deploy specialized agents for different tasks and monitor their progress from one unified dashboard.

The planet icon is a telling detail. The same icon previously appeared under the name "Orbit," a proactive assistant concept that was removed before launch, and TestingCatalog notes Hub Mode may be a revised version of that idea — though there is no confirmed connection beyond the shared icon.

The move fits Anthropic's push into multi-agent orchestration throughout 2026. Claude Opus 4.8 introduced Dynamic Workflows that let Claude orchestrate up to 1,000 sub-agents in parallel via a JavaScript orchestration script, but users only see the output, not the individual agents doing the work. Hub Mode could be the interface tying it together — the missing front end for a back end that already works.

The timing also aligns with how people actually use Claude. After Claude Cowork expanded to mobile and web in July, Anthropic said over 90% of Cowork sessions are not coding but business operations, content creation and workflow automation — workloads where a single chat thread becomes a bottleneck, and where a board that delegates different parts of a workflow to different agents at once would help.

Open questions remain. TestingCatalog explicitly notes there is no confirmed connection between the mobile toggle and the web Hub interface; they could be two entry points into the same system or two separate features sharing a name. Pricing and platform positioning are equally unknown — running multiple concurrent agents consumes significantly more resources than a single conversation, and whether Hub Mode lands on Max, Enterprise, or both will signal how Anthropic positions it.

For now, Hub Mode is an internal experiment with no release date or public documentation. But the pieces it connects — sub-agent orchestration, task boards, proactive capabilities — describe a product that would fundamentally change how users interact with Claude: instead of a conversation partner, Claude becomes a workforce manager.