Image: Anthropic / support.claude.com Anthropic Brings Claude Code and Claude Cowork to Government in FedRAMP High Beta
Claude Code and Claude Cowork are now in public beta for U.S. federal, state, and local agencies, running in a FedRAMP High authorized environment with hard spending caps and tamper-evident audit logs.
Anthropic launched a public beta of Claude Code and Claude Cowork inside Claude for Government Desktop on July 7, making both tools available to U.S. government agencies under a FedRAMP High authorized environment for the first time.
The offering covers federal, state, and local agencies. Claude Code is positioned for software teams modernizing legacy systems. Claude Cowork handles document work: memo drafts, RFP reviews, casework, and slide decks, working directly with files on the agency’s local machine.
What makes the government version different
The core difference from the commercial product is the governance layer. Administrators can set configuration defaults and allocate spending to sub-agencies while allowing each to manage its own users. Usage is purchased in fixed increments with hard spending caps, which lets agencies align AI usage with appropriated funds instead of risking overages on a credit card.
Conversation history stays on the agency-managed device. Inference runs inside Anthropic’s FedRAMP High infrastructure. Security teams get tamper-evident audit logs and documentation that can support an Authorization to Operate (ATO) process.
Anthropic also published its FedRAMP Secure Configuration Guide publicly. Penetration test summaries are available under NDA for agencies that need them during authorization reviews.
FedRAMP High matters here
FedRAMP High is the strictest certification tier for unclassified government data. It applies to systems handling sensitive but unclassified information across defense, health, and law enforcement contexts. Many agencies cannot legally use cloud AI tools that don’t meet this bar, which is why Anthropic’s earlier work getting Claude into AWS GovCloud and achieving DoD IL4/5 authorization through Amazon Bedrock was necessary groundwork for this announcement.
The government beta builds on that infrastructure rather than creating a separate environment.
Government agencies already using Claude
The announcement comes after several large deployments. The Department of Defense has a $200 million ceiling agreement with Anthropic. Lawrence Livermore National Laboratory has about 10,000 daily users on Claude. Washington D.C.’s Department of Health has also been using Claude for operational work.
The new Desktop beta extends access to agencies that want the more interactive Claude Code and Cowork experience rather than API integration.
Pricing
Agencies can choose standard seat pricing or define custom tiers with usage and model limits. There’s no single published price; each agency negotiates based on seats and usage caps. Eligible organizations can contact [email protected] for access.
The FedRAMP High public beta is available now. Broader rollout details haven’t been announced.
Sources: Anthropic — Bringing Claude Code and Claude Cowork to government, Anthropic — Expanded access across all three branches of government, Claude Help Center — Public Sector FAQs