Anthropic exits Claude Cowork research preview with enterprise features, launches Claude Managed Agents beta
Anthropic has promoted Claude Cowork from research preview to general availability, adding six enterprise features including role-based access controls, group spend limits, and usage analytics. The company simultaneously launched Claude Managed Agents in public beta—a composable API suite for building and deploying cloud-hosted agents without custom infrastructure work.
Anthropic Exits Claude Cowork Research Preview with Enterprise Features, Launches Claude Managed Agents
Anthropicis ending the research preview phase for Claude Cowork and introducing six enterprise-focused features as the tool transitions to general availability on macOS and Windows for all paid subscribers.
The six new enterprise capabilities include:
- Role-based access controls for permission management
- Group spend limits to cap usage costs across teams
- Usage analytics for consumption tracking
- Expanded OpenTelemetry support for observability integration
- Zoom MCP connector for workflow integration
- Per-tool connector controls for granular access management
Claude Cowork, which allows Claude to manage workflows directly on Mac systems, was introduced three months ago as a research preview. The general availability launch indicates Anthropic believes the tool is production-ready for enterprise deployment.
Claude Managed Agents Enters Public Beta
Simultaneously, Anthropic launched Claude Managed Agents in public beta—a suite of composable APIs for building and deploying cloud-hosted agents at scale. The tool abstracts away infrastructure complexity that traditionally required development teams to handle secure infrastructure provisioning, state management, permissioning logic, and agent loop reworking for each model upgrade.
According to Anthropic, Managed Agents pairs a performance-tuned agent harness with production infrastructure, reducing deployment timelines from months to days. Early users including Notion, Asana, and Sentry have already built solutions using the platform during preview phases. The tool is now available on the Claude Platform for all builders.
Context: Claude Mythos and Project Glasswing
These announcements follow Anthropic's recent introduction of Claude Mythos Preview and Project Glasswing. The company revealed this week that it is not publicly releasing Mythos—its announced most powerful model—but instead deploying it as a cybersecurity tool. According to Anthropic, companies including Apple are using Mythos to discover and resolve software vulnerabilities.
What This Means
Anthropics is signaling a shift toward infrastructure-as-a-service for agent deployment rather than competing solely on model leaderboards. By removing operational friction from agent development, the company aims to accelerate adoption among enterprises already committed to Claude. The enterprise feature set for Cowork—particularly spend controls and role-based access—directly addresses procurement concerns that typically slow AI tool adoption in regulated industries. The Managed Agents beta creates a moat around the Claude ecosystem by making it the cheapest path from prototype to production for agent-based applications.
Related Articles
OpenAI adds ChatGPT to Microsoft PowerPoint in public beta
OpenAI has integrated ChatGPT into Microsoft PowerPoint, allowing users to generate and edit presentation slides using natural language prompts. The feature is available in public beta to both free tier users and ChatGPT Business subscribers.
Anthropic's Unreleased Claude Mythos Preview Finds 10,000+ Vulnerabilities in One Month
Anthropic's unreleased Claude Mythos Preview model has discovered more than 10,000 vulnerabilities across partner organizations in its first month of deployment through Project Glasswing. The company reports partners are finding bugs at 10x their previous rate, with Cloudflare discovering 2,000 bugs and Mozilla finding 271 Firefox vulnerabilities — 10x more than with previous Claude models.
Anthropic Python SDK v0.104.0 adds thinking token count estimates for streaming responses
Anthropic released version 0.104.0 of its Python SDK on May 21, 2026. The update adds support for a thinking-token-count beta feature that provides estimated token counts in thinking block deltas when streaming responses from reasoning models.
AWS Launches Amazon Bedrock AgentCore for Deploying Production AI Agents
AWS has launched Amazon Bedrock AgentCore, a serverless runtime environment for deploying production AI agents. Turkish fulfillment company OPLOG demonstrated the platform's capabilities by building three business intelligence agents using Anthropic's Claude Sonnet, achieving a 35% reduction in sales cycles and 98% reduction in manual research time.
Comments
Loading...