Claude for Government bills agencies with a hard spend cap

Anthropic took Claude for Government out of public beta on September 30 and made it generally available to federal and state agencies. Pricing rests on a hard not-to-exceed cap. There are no seat fees, agencies buy usage in fixed increments, and spending can’t go past what an agency has obligated. Anthropic hasn’t said how large those increments are or what they cost.
The platform runs Claude’s coding and agentic tools inside a FedRAMP High authorized environment, according to Anthropic’s announcement on the Claude blog. FedRAMP High is the strictest tier of the federal cloud security authorization program. Anthropic also says agencies get capabilities comparable to what its commercial customers have, and that new features generally ship on the commercial release cadence.
The beta opened on July 7, when Claude Code and Claude Cowork came to Claude for Government Desktop. Anthropic’s July post said the desktop app is built on the same application commercial customers use. The rollout was presented as expanding Claude access across all three branches of the US federal government: executive, legislative and judicial.
The desktop app works directly with local files. Agency staff can use skills, plugins and projects for drafting memos, reviewing RFPs and handling casework, and Claude Code covers building and modernizing the software behind public services. The Claude Code command-line interface and Claude for Microsoft 365 are only now entering early access. They run in the same environment under the same admin controls, and agencies sign up through Anthropic’s public sector team. For developers in government, that leaves the terminal tool still a step behind the desktop app.
Administrators define user tiers with spend and model limits for each group, track usage by user and by model, and get burndown alerts before a balance runs low. The July post adds that program offices can tie AI spend to appropriated funds, either with standard seats or with seat tiers they define themselves.
Administration mirrors how departments are organized. A department-level administrator allocates prepaid usage to sub-agencies, and each sub-agency manages its own users. Agencies connect their own identity provider for single sign-on and can set it up themselves in the admin portal. SCIM group mappings then set rate limits, dollar caps and allowed models for each seat tier. SCIM is the standard identity providers use to sync users and groups into other systems. A layer of configuration sets defaults for sub-agencies, including what Claude can connect to and which features are switched on, and the July post says it also covers guiding instructions.
Every administrative action goes into an audit log that organization administrators can review in the product. The July post describes that log as hash-chained: each entry is cryptographically linked to the one before it, so an edit to an earlier record would show. On Anthropic’s side, sensitive operations need approval from two people. Usage exports contain only metering data, so agencies can answer ATO and inspector general requests without moving sensitive material. An ATO, or authority to operate, is the formal sign-off an agency official gives before a system goes into service, and Anthropic says its audit logs and documentation support that process.
Conversation history lives on the agency-managed device, while inference runs inside the FedRAMP High environment. The app deploys through agencies’ standard MDM (mobile device management) platforms. Anthropic’s FedRAMP Secure Configuration Guide is available through its trust center. The July post says there is also a penetration-test summary for the new desktop client, and that the formal change notification and pentest summary are available there under NDA.
Contracting officers can contract with Anthropic directly and award on general-availability terms. Anthropic stays the contracted and billing party, so agencies don’t need a separate cloud-provider relationship to get started. Agencies already on the platform can switch to the desktop application and bring their conversation history with them through an in-app import.
Up next
Barclays wants half its developers on Claude Code this year
“In the hands of most Barclays engineers by 2027” is how Anthropic describes the Claude rollout at the bank, and the first checkpoint comes sooner: 50% of its developers on Claude Code by the end of 2026.
Related stories
- Claude gets 40% off for every level of US government
- Pentagon to finish cutting Anthropic loose by month’s end
- Anthropic’s $150M buys Claude time for federal scientists
- Claude gets write access to Microsoft 365 on Dec 9
- Claude for Manufacturing plugin ships with 17 skills
- Claude Meetings joins Google Meet calls as a participant
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

