Anthropic has made Claude for Government generally available to U.S. federal and state agencies, running in an environment authorized at the FedRAMP High level, the company said in a blog post Wednesday. The release ends a public beta that began in July and comes with a usage-based pricing model that has no per-seat fees.

What Happened

The generally available package bundles the Claude desktop application, which agency staff can use to work with their own files, together with skills, plugins and projects for routine government work. Anthropic's examples include drafting memos, reviewing requests for proposals and handling casework. Claude Code is included so that public-sector development teams can build new software and modernize older systems.

Anthropic said new capabilities will generally reach the government offering on the same cadence as its commercial products, rather than trailing them by months. Two further products, the Claude Code command-line interface and Claude for Microsoft 365, are in early access and run in the same FedRAMP environment under the same administrative controls.

RelatedNew York City Comptroller Urges Anthropic to Delay IPO Over Safety and Governance Concerns

Existing beta customers can move to the desktop app and import their conversation history from inside the application. New agencies are directed to request access through Anthropic's government solutions page.

Why It Matters

Most of the announcement is about procurement and oversight rather than model capability, which reflects how agencies actually buy software. Instead of licensing seats, agencies purchase usage in fixed increments with hard spending caps. Administrators can define user tiers with their own spending and model limits, track consumption and receive burndown alerts as a block is used up. Department-level administrators can hand prepaid usage down to sub-agencies, each of which then manages its own users.

Access is tied to the agency's own identity provider through single sign-on, and SCIM group mappings set rate limits, dollar caps and permitted models for each tier. Anthropic said sensitive operations require approval from two people, administrative actions are written to an audit log, and usage exports contain only metering data. The company said that last point lets agencies respond to requests from authorizing officials and inspectors general without moving sensitive material. Conversation history stays on the agency-managed device, and the desktop app can be deployed through standard mobile device management tools.

Anthropic also said security teams receive documentation to support an agency's authority to operate (ATO), and that a FedRAMP Secure Configuration Guide is available through its trust center.

Context

FedRAMP is the federal program that authorizes cloud services for government use, and High is the most demanding of its three baselines, intended for sensitive unclassified data where a breach could have severe consequences. Clearing that bar is a prerequisite for many civilian agency deployments involving law enforcement, health or financial records.

The general release lands as frontier AI developers compete for government business and federal leadership pushes agencies to adopt AI tools. For a buyer, the hard caps and metering-only exports address two common objections to generative AI in government: unpredictable spending and uncertainty over where sensitive prompts and outputs end up.

Anthropic's post does not list which Claude models are available in the government environment, and it does not publish prices for the usage increments.

What to Watch

The next milestones are the move of the Claude Code CLI and Claude for Microsoft 365 out of early access, and whether Anthropic follows through on keeping government features in step with commercial releases. Also worth watching is whether agencies publicly name deployments now that the beta label is gone, and how the capped-usage model compares with the seat-based and enterprise-agreement terms offered by competing providers.