On April 9, 2026, Anthropic moved Claude Cowork — the desktop GUI that lets Claude work alongside you in a browser-style window — to general availability on macOS and Windows for every paid plan (Pro, Max, Team, Enterprise). The headline-grabbing feature inside it, Computer Use — the one that lets Claude move a cursor, click buttons, and type into your real desktop apps — remains a research preview restricted to Pro and Max subscribers, first added to Cowork on March 23.
The distinction matters. The wrapper is mature; the inside-the-app cursor-control feature is still being hardened. For technology leaders deciding what to roll out to a team this quarter, those are two very different commitments — and the recent press conflates them constantly.
What follows is an operator's read of what actually shipped this spring, where the technology pays back today, and the guardrails you need before any of it touches a real machine.
What Anthropic actually shipped
The spring 2026 release wasn't one announcement — it was a sequence of four. Tracking them in order is the easiest way to understand what's GA and what isn't.
- February 5 — Claude Opus 4.6. New top-tier model. Headline capability: longest task-completion horizon of any frontier model in production at release, plus an "agent team" structure and Claude inside PowerPoint.
- February 17 — Claude Sonnet 4.6. Promoted to default for Free, Pro, and Max users on Claude.ai. 1 million-token context window in beta. Around 78% on SWE-bench Verified. The OSWorld benchmark — the standard for desktop computer-use ability — moved noticeably with this release; early users reported "human-level capability" on tasks like navigating a complex spreadsheet or filling out a multi-step web form.
- March 23 — Computer Use joins Cowork (and Claude Code) on macOS. As a research preview for Pro and Max subscribers. Windows followed roughly ten days later. A companion Dispatch feature lets a phone send tasks to a desktop agent.
- April 9 — Cowork goes GA. Available across all paid plans on macOS and Windows, plus six enterprise features for larger deployments: role-based access control, group spend limits, usage analytics, expanded OpenTelemetry support, tightened connector permissions, and admin controls. Managed Agents also shipped the same day — background runs that don't require a foreground session.
- April 16 — Claude Opus 4.7. New top model, now generally available across the API, Bedrock, Vertex AI, and Microsoft Foundry.
If you're skimming the timeline, the operator-relevant takeaway is this: the desktop product is GA across paid plans; the inside-app cursor-control feature is preview-only and tier-limited. Treat them as separate decisions.
From chatbot to active worker
It's worth being precise about how long this curve has been building. Anthropic first introduced Computer Use on October 22, 2024, as a developer-only API toggle on Claude 3.5 Sonnet. The model could see a screenshot, infer what to click, and emit cursor coordinates and keystrokes that an external runtime executed. SWE-bench Verified scores jumped from 33.4% to 49.0% in that release — the first credible sign agentic coding could move out of demo videos.
For most of 2025, Computer Use stayed in that developer-API box. Real desktops were where the demos broke — agents would mis-click, lose track of state, and fall over on common UI patterns. The OSWorld scores climbing through the second half of 2025 told the story: the underlying model was getting steadily better, but it wasn't yet good enough to put in front of non-developers.
February 2026 changed the picture. Sonnet 4.6 brought a step-change on the desktop benchmark and 1M-token context. Cowork, which had launched in January as a research preview, suddenly had a model that could actually drive a spreadsheet without breaking. Two months later — April 9 — Cowork shed the preview label and shipped to every paid plan on both desktop OSes.
Computer Use itself is still on the trailing edge of that maturity curve. That's why it's gated to Pro/Max and still labelled research preview, even inside a now-GA product.
Where this pays back today
The honest answer is: high-frequency, low-judgement desktop work that already eats your team's attention. Concretely:
Tool-bridging. Most internal workflows already involve copying data between systems that don't have a clean API path. Pull something out of a CRM, reformat it, drop it into a spreadsheet, reconcile against a finance tool. The arithmetic against a $20-a-month Pro subscription is hard to argue with for a finance ops or sales ops team — particularly compared with a brittle RPA macro that breaks every time a UI updates.
Spreadsheet operations. A meaningful share of "AI for business" demos collapse the moment they meet a real spreadsheet — multiple sheets, conditional formatting, pivot tables that break when you sneeze. The OSWorld improvements in Sonnet 4.6 specifically targeted this category, and the difference is visible. Reconciling vendor invoices against a ledger is not glamorous, but it is exactly the kind of high-volume, high-error work that benefits from a tireless agent.
Research and brief assembly. Sub-agents based on Haiku 4.5 can collect and summarise inputs cheaply, while a single Sonnet 4.6 or Opus 4.7 instance handles the synthesis. We use this pattern in our own AI workflow automation builds: a router model dispatches narrow lookups to the cheapest model that can handle them, then a stronger model assembles the final answer. The cost profile of an agentic system is dominated by which model handles which step.
Runbook execution. New-hire setup, account provisioning, periodic compliance checks, scheduled report generation — anywhere a human is executing a procedure they have executed a hundred times before — is the natural target for an agent inside Cowork. The agent doesn't get bored, doesn't skip steps, and produces an auditable trail of what it did.
Background work via Managed Agents. The April 9 release added scheduled and queued runs that don't need a foreground session — overnight processing, recurring audits, "do this every Monday at 9 a.m." workflows. This is the production complement to the interactive Cowork experience. For automation teams, it removes the awkwardness of an agent that only runs when someone is at their machine.
What it does not replace, yet: anything where a wrong action is irreversible. Production deploys, financial trades, customer-facing communications without a human-in-the-loop. The model is good; it's not yet good enough to trust unsupervised on actions you can't take back.
What you have to get right before you ship
Treating any of this as a desktop pet is the fastest way to get burned. The teams putting it into production responsibly are doing four things.
Granular permissions, not a master switch. App-by-app approvals beat "let the agent do anything." When something fails — and it will — you want a tight blast radius and a clear log of exactly what was permitted at the time. Cowork's enterprise controls (RBAC, connector permissions) are a starting point, not a finished policy. The actual policy is yours to write.
A dedicated runtime. Putting an agent on a marketing manager's laptop, where it shares state with email and Slack and a half-loaded Chrome window, is asking for trouble. The teams doing this seriously now run Computer Use on a dedicated "AI workstation" — a secondary laptop or a remote VM — that the agent can take over without disrupting a human's working day. That is also where the audit trail lives.
Evaluation and drift monitoring. Computer-use agents fail in subtle ways. A UI changes, a button moves, the agent's chain of reasoning quietly drifts from "do X for the customer" to "do Y for a different customer entirely." You need ongoing evaluations against pinned scenarios and alerting when success rates dip. This is where agentic AI infrastructure discipline — model routing, eval harnesses, drift detectors — pays for itself many times over. Without it, you are shipping an agent into production blind.
A defined human escalation lane. State explicitly which decisions the agent makes alone, which it must request approval for, and how it surfaces ambiguous cases. Teams that skip this step end up either with an agent that does too much and earns a costly mistake, or one that does almost nothing because every case is technically "ambiguous."
The competitive picture for incumbents
The clearest casualty of this category is the legacy RPA stack. Tools that charge tens of thousands of dollars per year to record macros against UI selectors are now competing against a $20-a-month subscription that reasons about what is on the screen. Microsoft and Google will respond — Copilot and Gemini will acquire similar native capabilities at the OS level over the next year, which will compress the differentiation again. The window for first-mover advantage on internal automation is open, but it is not infinite.
For technology leaders in Southeast Asia specifically, two things matter beyond the raw capability. Latency to the model favours regions with strong Anthropic, Bedrock, and Vertex coverage, and a thoughtful governance posture is non-negotiable because data-residency rules in financial services and healthcare do not relax just because the agent happens to be running on a desktop.
How we're approaching it for clients
We treat Cowork-plus-Computer-Use the same way we treat any other production component: scope, instrument, evaluate, ship behind a flag, monitor. The difference with desktop agents is that the failure modes are louder — a mis-click on the wrong record is far more visible than a regression in a backend service — so the bar for evaluation infrastructure is higher, not lower.
The pattern we keep returning to is hybrid. An agent handles the deterministic, high-volume work that used to consume a person's morning. A human reviews exception cases and approves any action with a real-world side effect. The whole system runs on instrumented infrastructure that catches drift before it becomes an incident. That is not a slide. It is the only configuration we have seen actually hold up at the scale where it matters.
Cowork going GA isn't the finish line of agentic AI. It is the part of the curve where the wrapper around the agent stops being experimental — even if the agent inside it is still being tightened. That makes April 9 the right moment for operators to stop watching from the sidelines and start scoping their first contained pilot.
