Anthropic Opens Claude for Government to Agencies With Hard Spending Caps
The government service pairs coding and office work with department-level controls and direct contracting. Claude Code CLI and Microsoft 365 access remain in early access.
Loading page…
The government service pairs coding and office work with department-level controls and direct contracting. Claude Code CLI and Microsoft 365 access remain in early access.
Listen to this story
Anthropic made Claude for Government generally available to U.S. federal and state agencies on September 30, ending its July public beta and enabling direct contracts without a separate cloud-provider relationship. The FedRAMP High environment combines desktop access with prepaid usage and hard spending ceilings; administrators can allocate budgets across sub-agencies and limit model access. Claude Code’s command-line interface and Claude for Microsoft 365 remain in early access, and the civilian rollout does not lift Pentagon restrictions tied to Anthropic’s dispute over surveillance and autonomous-weapons safeguards.
Agencies pay by usage in fixed increments, with no seat fees and not-to-exceed caps designed to keep charges within obligated funds.
Conversation history stays on agency-managed devices; administrative actions are logged, while usage exports contain metering data rather than conversation content.
Sensitive operations on Anthropic’s side require two-person approval, and the company provides documentation for agencies’ authorization-to-operate process.
U.S. federal and state agencies can now contract for Claude for Government on general-availability terms. Anthropic opened the service on September 30, 2026, moving it beyond a public beta that began in July. The rollout brings Claude’s coding and multistep task capabilities into a FedRAMP High-authorized environment, with controls over spending, model access and administration.
The change is also contractual: procurement officers can now award contracts directly to Anthropic on general-availability terms. Agencies do not need a separate relationship with a cloud provider to get started. The company says the service offers capabilities comparable to its commercial products, with new features generally arriving on the commercial release schedule.
Claude works with files on the desktop. Anthropic lists memo writing, reviews of requests for proposals and casework among the intended uses, supported by skills, plugins and projects. It also positions Claude Code as a way for public-sector teams to build and modernize the software behind public services. Those are the company’s proposed uses, not measured agency outcomes.
Not every part of the announcement has the same availability status. The Claude Code command-line interface and Claude for Microsoft 365 are rolling out in early access. Both use the same government environment and administrative controls. Agencies seeking those tools must contact Anthropic’s public-sector team rather than treat them as generally available.
Anthropic says there are no seat fees. Agencies buy usage in fixed increments, with a hard not-to-exceed cap designed to keep charges within the money the agency has obligated. Administrators can track consumption by user and model, and receive alerts before the remaining balance runs low. The billing model pairs usage charges with an explicit budget ceiling.
The controls also follow an agency’s departmental structure. A department can allocate prepaid usage to sub-agencies while letting each manage its own users. That allocation sits alongside separate limits on which models people can use and how much they can spend. Anthropic describes three layers of administrative control:
Sensitive operations on Anthropic’s side require two-person approval. Security teams and agency officials also receive documentation supporting the agency’s authorization-to-operate process. Anthropic says the metering-only exports let agencies answer authorization and inspector-general requests without moving sensitive material. The records are designed to support oversight without exporting conversation content as usage data.
Deployment uses agencies’ standard device-management platforms. Existing customers can move to the desktop application and import their conversation history through the app. Anthropic also provides a secure-configuration guide through its trust center. These are the rollout’s practical setup steps, alongside the direct-contracting route for new agencies.
The civilian rollout comes against a different backdrop at the Pentagon. The Decoder reports that the department banned Claude and designated Anthropic a supply-chain risk in March after the company refused to remove restrictions on mass surveillance and autonomous weapons.
According to the outlet, a San Francisco judge struck down one of two designations as unlawful retaliation in late August. A Washington appeals court upheld the other in late September. The September 30 product announcement establishes general availability for federal and state agencies; it is not an announcement that the Pentagon’s restriction has ended.
Loading discussion...
Join the conversation
Explain which approach you think better protects public money.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.