Anthropic Brings Claude Code and Cowork to the US Government

Anthropic announced on July 7 that it is rolling out Claude Code and Claude Cowork in public beta for the US federal government. The tools run inside Claude for Government Desktop, the same application used by commercial customers, but operating in an environment authorized under FedRAMP High.

What agencies get

Claude Code is meant to let public-sector teams build and modernize the IT systems behind government services. Claude Cowork works directly on files on the user's computer, letting agency staff ask the model to draft memos, analyze requests for proposals (RFPs), manage casework, or build presentations.

Anthropic says the offering pairs agentic coding and office-work tools with controls tailored to government use. Conversation history is stored locally on agency-managed devices, and the package includes department-level administration, spending and model limits, tamper-resistant audit logs, and documentation to support the Authority to Operate (ATO) process.

A FedRAMP High environment

The company published a publicly available FedRAMP Secure Configuration Guide and issued a formal notification of change with technical details. A penetration-testing summary is available on request, under a non-disclosure agreement, through Anthropic's trust center. The app installs through the standard MDM platforms government agencies already use, meant to make deployment easier for IT teams without additional infrastructure.

The billing model is based on standard or custom seat packages with spending and model limits, purchased in fixed increments with a hard cap, and per-user, per-model usage tracking in the admin console. Automatic budget-depletion alerts are meant to limit the risk of runaway costs in public bodies, where budgets are tightly planned.

Cowork comes to phones and browsers

Alongside the government version, Anthropic extended Claude Cowork beyond the desktop app to web browsers and mobile devices for all customers. The beta launched July 7, first for Max plan subscribers, with a plan to roll it out to other subscription tiers in the coming weeks.

"Works through the desktop app while it is open" - Anthropic, describing how Cowork functions on mobile and in the browser

The new architecture lets users start a task on one device and continue it on another. Someone can check progress, approve next steps, or resume a session remotely from a phone or browser. The caveat matters: access to local files and control over the computer require the desktop app to keep running in the background, so the mobility applies to overseeing a task, not full independence from the computer.

What it means for Poland

For Polish companies and public institutions, the direct relevance is mainly the precedent: Anthropic is showing a ready-made pattern for deploying AI agents in environments with strict security and audit requirements, similar to those in place under EU frameworks for the public sector and regulated industries. Data on how Cowork is actually used, dominated by office and operational tasks rather than coding itself, confirms that AI agents are entering administrative work faster than the coding-focused narrative so far would suggest.

Competition between Anthropic and Microsoft over the market for tools that hand office work to AI agents is intensifying, with both companies launching similar features within weeks of each other. For IT departments considering such tools, that means more choice, but also a need to compare billing models, the scope of administrative controls, and data-retention requirements before picking a vendor.

Sources: Bringing Claude Code and Claude Cowork to government (claude.com), Anthropic Opens Claude Cowork Beta to Mobile and Web (winbuzzer.com)