Anthropic announces Claude Managed Agents with multi-agent orchestration and 'Dreaming' self-improvement feature; API volume up 17x year-on-year

Anthropic launched three new Claude Managed Agents features: multi-agent orchestration (for coordinating fleets of agents on complex tasks), Outcomes (to define success criteria so Claude iterates to achieve them), and Dreaming (a research preview where Claude inspects previous sessions overnight to self-improve and create new memories). Multi-agent orchestration and Outcomes are in public beta; Dreaming requires access request.

The company is doubling the five-hour limit for Claude Code for Pro, Max, and Enterprise customers, and announced a partnership with SpaceX to use capacity from their Colossus data center.

API volume on the Anthropic platform is up 17x year-on-year. One customer (eve) achieved 'frontier model quality at 5x lower cost' by having Sonnet call Opus as an advisor, generating both higher benchmark results and lower costs.