Anthropic Claude Platform build artifacts blog post graphic Image: Anthropic / claude.com
by Michael Joiner

Anthropic Ships Computer Use, the Browser Tool, the Skills API, and the Files API as Production-Ready

The four features move out of beta and into general availability on the Claude Platform on August 20, with batch multi-action turns, HIPAA eligibility, 5x higher Files API rate limits, and 1 TB storage per organization.

Share

Anthropic moved four Claude Platform features from beta to general availability on August 20, 2026: computer use, a new browser use tool, the Skills API, and the Files API. The combination is meant to cover the full stack for building agents that operate software, apply custom knowledge, and produce finished files.

Computer Use

The computer use tool is now available without a beta header as computer_toolset_20260801. The two meaningful changes from the beta:

Multi-action turns. The agent can now execute several operations in one turn instead of a single click or keystroke at a time. This reduces round trips for tasks that need sequences of actions.

HIPAA eligibility. Computer use is now covered under Anthropic’s Business Associate Agreement, which matters for teams building agents that handle regulated health data.

One early customer, Davide Locatelli from a claims processing team, described what this looks like in practice: “On the new computer use tool, our longest claims workflow went from 32 minutes to 13, cost per task fell about 30%, and completion hit 100%.”

Browser Use Tool

The browser use tool is a new addition that ships alongside the computer use update. Where computer use works from screenshots and pixel positions, the browser tool reads page structure and targets web elements directly. It’s more reliable for agents that need to interact with web applications, particularly forms and structured interfaces where pixel-based targeting gets brittle.

Skills API

The Skills API lets you give Claude your team’s expertise packaged as versioned skill sets: instructions, scripts, and templates that load only when a task calls for them. Skills run in Claude’s sandbox environment, which keeps execution isolated without requiring you to build your own infrastructure for it.

The versioning model is the key operational detail. You can update a skill and roll it back if something breaks, which isn’t possible when that knowledge is just embedded in a system prompt.

Files API

The Files API adds three improvements:

  • Automatic expiration via expires_in_seconds, so files clean up without manual deletion
  • 5x higher rate limits (up to 500 requests per minute)
  • 1 TB storage per organization

The practical use case is the one the API was always designed for: upload a file once, get back a file_id, and reference that ID across many agent requests without re-uploading the content each time.

How They Fit Together

Anthropic frames these four as a production agent stack rather than separate features. Computer use handles operating software with no API. The browser tool handles web apps. Skills carry your team’s methodology. Files carry your documents. Agents built from this combination can complete tasks end-to-end without needing custom infrastructure at each layer.

All four are available in the Claude API now. No beta headers required.

Sources:

Share