Claude Cowork Exits Preview — Now GA Across All Paid Plans With Enterprise Controls
Anthropic’s autonomous desktop assistant Claude Cowork has dropped the “research preview” label and is now generally available on macOS and Windows across all paid subscription tiers. The GA release brings a suite of enterprise controls: role-based access via SCIM integration with identity providers, spend limits and usage governance, expanded analytics dashboards, and OpenTelemetry support for observability pipelines.
The timing is telling. Non-engineering teams — operations, marketing, finance, and legal — now account for the majority of Cowork usage among early enterprise adopters. Anthropic is betting that the tool that started as a developer-adjacent experiment is ready for company-wide rollout.
Managed Agents Early Adopters Go Live — Notion, Rakuten, Asana Deploying at Scale
Less than a week after Claude Managed Agents launched in public beta, early adopters are sharing production results. Notion is running parallel task agents inside workspaces. Rakuten deployed enterprise agents across five departments within a week each. Asana built “AI Teammates” that work alongside humans in project management workflows. Sentry and Vibecode round out the launch cohort.
Managed Agents costs standard Claude API token rates plus $0.08 per session-hour. The platform handles sandboxed code execution, checkpointing, credential management, scoped permissions, and end-to-end tracing — the infrastructure grunt work that previously kept teams from shipping agents to production.
Claude Code 2.1.101 Ships Team Onboarding, Enterprise TLS, and Remote-Session Fixes
The latest Claude Code release adds a /team-onboarding command that generates a teammate ramp-up guide from your local usage patterns. Enterprise users get automatic trust for OS CA certificate stores, which means TLS-inspecting proxies work out of the box without manual cert configuration (set CLAUDE_CODE_CERT_STORE=bundled to opt out). Remote-session setup via /ultraplan now auto-creates a default cloud environment instead of requiring web setup first.
Bug fixes include Homebrew cask update prompts, ctrl+e multiline jumps, fullscreen scrolling, and idle-return token count accuracy. The Write tool’s diff computation is also faster for large files.
Claude Haiku 3 Retires April 19 — One Week Left
Seven days remain before claude-3-haiku-20240307 goes dark. After April 19, API calls to the original budget model will fail. The migration path is Claude Haiku 4.5 (claude-haiku-4-5-20251001), which is faster, more capable, and priced at $0.80/$4.00 per million tokens. If you haven’t tested against Haiku 4.5 yet, this weekend is your last comfortable window.
“Claude Mania” Dominates HumanX — Anthropic Steals the Spotlight From OpenAI
CNBC’s vibe check from the HumanX conference in San Francisco says it all: Anthropic is the talk of the town. Among 6,500 executives, founders, and investors, Claude Code was the tool on everyone’s lips. Glean CEO Arvind Jain said Claude Code has inspired “Claude Mania,” putting pressure on business leaders to deploy it. Bloomberg’s coverage confirmed that OpenAI no longer dominates the conversation at the industry’s main events — for now, that distinction belongs to Anthropic.
The Mythos announcement earlier in the week only amplified the buzz, with attendees debating the implications of a model that can find zero-days in every major OS even as its rollout remains limited to 50 vetted partners.
Taipei Times: Mythos Alarm Bells — Fair Warning or Marketing Trick?
A Taipei Times opinion piece asks whether Anthropic’s dramatic Mythos disclosure is genuine safety transparency or a calculated marketing play. The argument: by revealing Mythos’s cyber capabilities in a controlled, headline-grabbing way, Anthropic positions itself as the responsible AI company while generating the kind of press money can’t buy. The piece concedes that the restricted-access approach is sound but questions whether the “alarm bell” framing was calibrated for maximum media impact rather than minimum risk.
It is a fair question, and one the industry will keep asking as frontier labs navigate the tension between transparency and theatrics.
Cowork GA Is Anthropic’s Bet That AI Agents Are Ready for the Whole Company
The most interesting data point in the Cowork GA announcement is not the RBAC or the OpenTelemetry hooks. It is that non-engineering teams already represent the majority of usage among early enterprise adopters. Marketing, ops, finance, legal — the people who were supposed to be years away from using AI agents are already the primary users.
That shifts the competitive frame entirely. This is not Claude Code for non-coders. This is a new product category: the autonomous desktop assistant that works across your files, tools, and workflows without requiring you to write a single line of code. If Cowork can hold that position as enterprise deployments scale, it becomes the front door to Anthropic for the 90% of knowledge workers who will never open a terminal. The enterprise controls shipped this week are table stakes for that bet. The real question is whether the agent can be reliable enough, across enough workflows, to justify company-wide seats.