Anthropic makes computer use, Skills API, and Files API generally available

The short version

Anthropic says computer use, the new browser use tool, the Skills API, and the Files API are now generally available on the Claude Platform. The announcement is a meaningful shift from isolated demonstrations to a set of production components for agents that can operate software, use a team's procedures, and return files. Anthropic's announcement is the source for the availability and feature details.

Computer use lets an agent click, type, and scroll in software from screenshots. Browser use adds page structure so an agent can target a web element instead of relying only on coordinates. The Skills API lets teams upload and version instructions, scripts, and templates. The Files API stores documents an agent reads or creates.

What changed

Computer use adds browser use and multi-action turns

The updated computer use tool can take several actions in one turn instead of one action per model call, according to Anthropic. Anthropic also introduced a browser use tool that combines screenshots with page structure. This is designed for web applications where an agent must identify a field or button, not just click a fixed position. See Anthropic's product announcement.

Anthropic says computer use is also eligible for HIPAA-regulated workloads under a business associate agreement. Teams still need to evaluate their own security, authorization, and audit requirements before using the feature with regulated data. The announcement states the product scope, not an approval for every workflow.

Skills API turns procedures into reusable packages

A skill is a folder containing instructions, scripts, and templates. The Skills API lets a team upload and version those materials, then attach them to a request. Anthropic says the skills run in Claude's code execution sandbox, so the team does not need to host the skill itself. The official announcement describes the Skills API.

This matters for organizations where a process is currently known by one person. A claims procedure, a credit memo format, or a review checklist can be stored as an explicit package. The package still needs ownership, review, and change control. Turning a procedure into a skill does not by itself make the procedure correct.

Files API gives agents durable document references

The Files API provides storage for documents an agent reads and writes. Anthropic says a PDF or spreadsheet can be uploaded once and referenced by ID in later requests, rather than sent again each time. The general availability announcement also lists automatic file expiration, five times higher rate limits, and 1 TB of storage per organization. These limits and capabilities come from Anthropic's announcement.

For teams, the operational question is which files should persist, how long they should remain available, and who can reference them. File retention should be part of the workflow design rather than an afterthought.

The workflow Anthropic is describing

Anthropic uses a claims workflow as an example: an agent reads an intake document from the Files API, follows a skill containing the team's filing procedure, completes a submission in an insurer's web portal with browser use, and saves the confirmation as a file. This example is described in the official announcement.

The important change is the connection between four layers:

1. A document gives the agent the case context.
2. A skill gives it the team's method.
3. Computer or browser use lets it operate software.
4. The Files API preserves the output for later review.

That combination is more significant than any one tool. It gives developers a way to model an end-to-end task instead of stopping at text generation.

Availability, pricing, and platform support

Anthropic says computer use, browser use, the Skills API, and the Files API are available on the Claude Platform. The Skills API and Files API are also available through Microsoft Foundry. The updated computer use and browser use tools are coming soon to Google Cloud's Vertex AI, according to the announcement. Existing beta integrations continue to work while customers migrate. Anthropic's availability statement is the current source for these details.

The announcement does not include a standalone price table for these capabilities. Teams should check their Claude Platform, Microsoft Foundry, or cloud contract for current model, tool, storage, and usage charges before estimating a production budget.

How this compares with other agent platforms

The market is moving toward agents that can do more than answer a question, but the boundaries differ by product. Claude's update is developer-facing: it supplies APIs and tools that a team can compose into its own application. Consumer assistants and office-suite copilots usually optimize for a ready-made interaction surface. An AI-native workspace such as Kylon places the review conversation, permissions, connected tools, and recurring workflows around the agent's work.

The comparison is not about which product has the longest feature list. It is about where context and accountability live:

Claude's tools are useful when a development team wants to build a custom agent application. Kylon is useful when the team wants the work, review, and institutional context to remain visible to the people responsible for it.

What teams should test first

1. Start with a non-sensitive document and a reversible browser task.
2. Keep computer use and browser use behind explicit human approval.
3. Version the skill and assign an owner before connecting it to production data.
4. Define file expiration, access rules, and the record that proves completion.
5. Measure corrections, blocked actions, time to completion, and handoffs.
6. Expand permissions only after the review path works with real users.

Bottom line

Anthropic's general availability release brings computer use, browser use, the Skills API, and the Files API together as a production toolkit for agents. The opportunity is clear: an agent can combine documents, procedures, software interaction, and finished files in one workflow. The difficult part is equally clear: teams must design permissions, review, retention, and ownership around that capability. Read the full announcement from Anthropic.