Skip to main content
Who this is for: Security, compliance, and IT reviewers who are assessing Claude for Government for their agency, and administrators who need to explain the product’s runtime behavior.
The answers on this page address the security and data-handling questions that come up most often during agency security review, for the Claude Desktop application and for the standalone Claude Code command-line tool, each in its own section. For assurance materials such as the security architecture overview, SOC 2 Type 2 report, and penetration testing summary, request access through the Anthropic Trust Center.
These cases are covered on other pages:

Claude Desktop

The sections below cover the Claude Desktop application, which offers three ways to work with Claude: Chat for simple conversations, Cowork for longer tasks with a local workspace folder, and Code for software development. With the Standard home, a message sent from the home screen starts a Cowork session on a device that can run Cowork.

Sandbox and isolation

In Cowork, and for the file-analysis steps in Chat, the Claude Desktop application runs shell commands and model-written code inside a dedicated local virtual machine. In Claude for Government, this sandbox is always the execution path for the code and shell commands that Claude runs in Chat and Cowork. The agent loop, built-in file tools, web fetch, and the connector client run in the Claude Desktop application on the user’s device. Code sessions run on the workstation itself rather than in the virtual machine, as described under Code in Claude Desktop. For the detailed threat model and isolation design, request the security architecture overview through the Anthropic Trust Center. Shell commands that Claude runs in the sandbox work on the folders the user has attached, a scratch area, and read-only reference material (such as skill and plugin directories). Shell commands cannot work on the rest of the user’s files.
  • Attached folders. When a user attaches a local folder to a Cowork session, the entire folder is made available to that session’s sandbox as a filesystem mount, so changes Claude makes are written directly to the folder on disk. For mapped drives, see Network drives on Windows.
  • Attached files. Files the user attaches individually to a conversation are copied or hard-linked into a per-conversation uploads directory and mounted read-only; where the filesystem hard-links, edits to the original file while the conversation is open can be visible to it.
  • Folder restrictions. Administrators can restrict which local folders users may attach with Allowed workspace folders on the Config page. The desktop client then refuses folders outside that list in the workspace picker, in requests Claude makes during a session, and in Claude’s file tools. In Cowork, Claude’s file tools may read and write in the listed folders. See Desktop and filesystem access for how folder scoping is enforced.
  • File analysis in Chat. Some attachment types, such as Excel and PowerPoint, need a conversion step that runs inside the sandbox. See Advanced file analysis in Chat. You do not need to make Cowork available to get this capability. See Chat in Claude Desktop for how Chat handles file attachments.

Code in Claude Desktop

Code sessions use Claude Code built into the desktop application and run on the user’s workstation itself, not in the virtual machine. The shell commands Claude runs during a Code session execute on the workstation’s own operating system under the user’s own account. On macOS and Linux, those shell commands run inside an operating-system-level sandbox that the application builds from your organization’s Allowed network hosts and Allowed workspace folders settings on the Config page. The sandbox is in place whenever either setting restricts access, which the default configuration does. Inside the sandbox, a command can create or change files in the session’s folder, the other folders Allowed workspace folders permits, and temporary locations. Unless Block reads outside workspace folders (Claude Desktop 1.46388.1 or later) is on, it can read any file on the device that the user’s account can open. By default it runs without asking the user first. A user can exempt specific commands from this sandbox in a Claude Code settings file, and an exempted command runs outside the sandbox under the permission mode the user selects for the session. In their own Claude Code settings, a user can also widen which files sandboxed commands can read or change. The sandbox limits Claude, not a user who chooses to loosen it.
  • Windows. On Windows, there is no operating-system-level sandbox for Code sessions. Shell commands run directly on the device under the permission mode the user selects for the session and under your agency’s own endpoint and network controls. The Allowed network hosts and Allowed workspace folders settings do not confine what those commands can reach, read, or change.
  • Linux. On Linux, the sandbox requires the bubblewrap and socat packages, so install both on each workstation as described under Set up Linux and WSL2 in the Claude Code documentation. If either package is missing or the sandbox cannot start, shell commands can run outside it, as they do on Windows.
  • Container workspaces. In container-based Linux environments, such as cloud development workspaces, the sandbox can fail to start, and shell commands in Code sessions can then fail with a bwrap error. To run Code sessions there, set Shell command sandbox to Off under Advanced settings in the Claude Code section of the Config page, at the level that applies to the members who work in those environments, such as a directory group. Off applies to those members on every device they use, not only in the container workspaces. It reaches their Code sessions where your organization’s Claude Code settings do. Shell commands in those Code sessions then run without the sandbox, under the permission mode the user selects for the session, as they do on Windows, and the Allowed network hosts and Allowed workspace folders settings no longer confine what those commands can reach or change. Off also turns the sandbox off in those members’ sessions in the Claude Code command-line tool.
  • The terminal. Claude’s terminal tools type commands into the Code tab’s terminal, which is the user’s own shell (PowerShell on Windows). The app asks the user to approve each of those commands by default, and they run as the user, outside any sandbox. On macOS and Linux, Claude Desktop 4.23118.0 or later offers these tools in local sessions only where neither Allowed network hosts nor Allowed workspace folders restricts access. Administrators can turn these terminal tools off with the Terminal panel tools card on the Config page. The Blocked shell commands list does not block commands Claude types into the terminal, but the Require approval for each command sub-setting on the Shell commands card applies to them in every permission mode.
  • Folders and approval. On every operating system, the application starts a Code session only in a folder that Allowed workspace folders permits when that setting is configured, and Claude’s file reading and editing tools then work only inside the permitted folders. Administrators can also require a prompt before Claude uses its shell tool, with the Require approval for each command sub-setting on the Shell commands card.
  • Remote and WSL sessions. Code sessions in Claude for Government run on the local workstation by default. SSH remote sessions are off unless your agency allows specific hosts in Claude Code’s managed settings on the device; see Interaction with Claude Code managed settings on the device. On Windows devices set up through device management, Code sessions inside a Windows Subsystem for Linux (WSL) distribution are off by default.
  • Local sockets. A command that talks to a local agent through a socket, such as git signing a commit with a key held in an SSH agent or a hardware-backed key manager, cannot reach that agent from inside the sandbox. On macOS, a user who needs such a command to work can list the agent’s socket path under sandbox.network.allowUnixSockets in their Claude Code settings file, which keeps the command inside the sandbox. Every sandboxed command in that user’s Code sessions can then ask the agent to sign or authenticate. A sandboxed command can change files that git runs during a commit, such as hook scripts kept in the working tree, and a git commit exempted with sandbox.excludedCommands runs that code outside the sandbox. See Sandbox settings in the Claude Code documentation.
See how your configuration reaches Code sessions. If your agency also deploys Claude Code’s own managed settings to the same devices:

Network egress, required domains, and proxies

You manage the egress allowlist with the Allowed network hosts setting on the Config page. For configuration and model inference, the application reaches the Claude for Government service hostname provided to your agency during onboarding. Sign-in happens in the user’s default browser, which must reach that same hostname, the Claude for Government sign-in service (a separate host that your Anthropic representative provides), and your agency’s identity provider. See the network prerequisites in Connect Claude Desktop to Claude for Government. Claude for Government does not publish IP addresses for its service and sign-in hosts. The IP addresses page in the Claude API documentation covers the Claude API, not the Claude for Government hosts. Anthropic-bound telemetry endpoints are not contacted in Claude for Government. The app also reaches the hosts listed under Non-essential services in Required egress paths. If traffic is blocked, the app will run without icons, previews, and widgets.
  • Does Claude for Government depend on claude.ai or anthropic.com? Sign-in, configuration, and model inference do not: none of the hosts they use are under either domain. Blocking *.claude.ai and *.anthropic.com leaves sign-in and inference working. Blocking *.claude.ai also blocks downloads.claude.ai, which prevents Chat conversations, Cowork tasks, and Code sessions from starting on devices installed with the standard installer unless the app has already downloaded the components they need from that host. App updates often change one or both of those components, and the sandbox virtual machine that runs shell commands in Cowork and Advanced file analysis in Chat then cannot start until the app has downloaded the new versions from that host. Devices installed with the offline installer variant, which includes those components, are not affected. On Claude Desktop versions earlier than 1.37937.0, web fetch in Code sessions first sends the page’s hostname to api.anthropic.com and fails where that host is blocked. Automatic application updates use hosts under these domains, so an agency that blocks these domains distributes updates itself, as described under Automatic updates. The hosts are listed under Auto-updates in Required egress paths. See Block automatic updates and Restart deadline for updates. A personal Claude account cannot sign in to Claude for Government, and a Claude for Government account cannot sign in to claude.ai, so there is no shared sign-in surface to restrict.
  • Egress allowlist. The allowlist governs outbound network access from the shell commands and package installs of Cowork sessions, which run in the sandbox virtual machine, from the sandboxed shell commands of Code sessions on macOS and Linux (see Code in Claude Desktop), and from the host-side web fetch tool in Chat and Cowork. It does not govern web search (which routes through the Claude for Government service) or connector traffic (covered under Connectors below). When the list is empty or unset, package installs and page fetches from those tools fail. On Claude Desktop versions earlier than 2.37337.0, those tools can still reach the Claude for Government service address and, if you have set a Telemetry endpoint on the Config page, that collector’s host. The list accepts exact hostnames, wildcard patterns such as *.example.com, or * to allow all outbound traffic. See Web search and web fetch for the full allowlist semantics.
  • Proxies. The desktop application follows the operating system’s proxy settings, including PAC URLs. Where no proxy settings reach the Cowork sandbox, as on Linux, its commands connect directly. See Network proxy for the traffic that bypasses the proxy. TLS inspection at your proxy should work; validate this in your environment before rollout. Web search requests pass through your proxy to the Claude for Government service, and the service’s onward call to the search provider originates from inside the FedRAMP High boundary.

Approvals and Auto mode

By default, Claude for Government prompts the user for connector actions, for each web search, and, in Cowork, when Claude asks to add another folder to the session. In Chat and Cowork, Claude’s file tools do not write outside the attached folders, the session’s working folder, memory, and the folders listed in Allowed workspace folders. In Chat and Cowork, shell commands run without a prompt because they run inside the sandbox virtual machine. Web page fetches run without a prompt in both Chat and Cowork and are checked against the egress allowlist. For most tools, users who see an approval prompt can choose Always allow, which suppresses that prompt for them going forward. Turning on the Require approval for each search, Require approval for each fetch, or Require approval for each command sub-setting on the Web search, Web fetch, or Shell commands card stops Always allow from suppressing that tool’s prompts. In Chat on Claude Desktop versions earlier than 2.110.0, every shell command prompts regardless. By default, members cannot switch to Auto mode, which Cowork shows as Automatically approve. When an administrator sets the Auto mode setting to Allowed on the Config page, Claude Desktop offers it in Cowork and in Code sessions, and members can switch the Claude Code command-line tool to it. New Code sessions in Claude Desktop 2.9939.2 or later and new interactive sessions in the Claude Code command-line tool 2.1.283 or later then start in it on models that support Auto mode, unless another mode is chosen. Cowork does not offer a bypass-permissions mode. For Code sessions and the command-line tool, see Bypass-permissions mode.
  • Connectors an administrator adds. For each connector an administrator adds, the administrator can switch individual tools on or off under Tool policy. A tool that is on stays available and each user approves every use. A tool that is off is blocked. Users cannot loosen these settings. Tools that are not listed in a connector’s Tool policy, including tools the server adds later, stay under each user’s control.
  • Microsoft 365 connector. The Microsoft 365 connector always asks for approval before write actions such as sending mail, changing calendar events, or posting Teams messages, and no Claude for Government setting removes that prompt. The Microsoft 365 card has no per-tool setting, and the connector’s read tools follow each user’s own approval choices.
  • Can individual shell commands be allowlisted enterprise-wide? Not in Chat or Cowork. Administrators can turn the built-in Web search, Web fetch, and Shell commands tools on or off, or require approval for them, but cannot allowlist individual commands within those tools. In Code sessions, the Claude Code section has no list of allowed commands. Permission rules in a Claude Code managed-settings file on the device can allow individual commands, but they apply in Code sessions only where your organization’s Claude Code settings don’t reach them, as Code in Claude Desktop describes.
  • Can Auto mode be disabled when a sensitive connector is attached? Not for one connector alone. To limit a sensitive connector while Auto mode is allowed, switch the connector’s tools off under Tool policy on the Config page.
See Connector tool policies for members, Tool and connector cards, and Connectors for the available settings.

Connectors

Connectors are called from the desktop application, outside the sandbox. The built-in Microsoft 365 connector and administrator-added connectors call their endpoints directly from the user’s device, through the system proxy where one is configured. OAuth tokens for both are stored encrypted on each user’s device using operating system encryption (macOS Keychain on Mac, DPAPI on Windows). A user’s sign-in to a server that a plugin declares is stored differently, and on Windows and Linux it is not encrypted: see Credentials. For administrator-added connectors, the bearer header entered on the Config page is delivered to each user’s desktop. The Let members add their own connectors switch on the Config page controls whether end users can add local MCP servers of their own in Claude Desktop, under Settings, then Developer. It is off by default, as described under Let members add their own connectors. The Let members add plugin marketplaces and Let members add their own plugins switches on the Config page are off by default. Both need Claude Desktop 1.37937.0 or later; on earlier versions members can add marketplaces and plugins. Hook commands and local MCP servers run as programs under the user’s own account, outside both the sandbox virtual machine and the Code session sandbox. For the plugins an administrator adds, see Plugins that run code. Artifacts follow the same tool settings as Claude’s direct connector calls: a connector tool that an administrator has switched off is refused, and a tool that requires approval on every use never runs from an artifact without the user’s approval of that call.

Telemetry and logging

In Claude for Government, Anthropic-bound error and usage telemetry is always disabled. The OpenTelemetry export to your own collector is a separate setting and sends data only to the endpoint you configure. You set the OpenTelemetry endpoint with Telemetry endpoint on the Config page. The Telemetry content capture setting on the Config page adds prompt, response, and tool content to the export for the categories you select, and nothing is selected by default. Is there an inline DLP or inspection point? No. Claude for Government does not include an inline content-inspection or DLP gate. Claude for Government doesn’t support routing Claude Desktop through an LLM gateway that your agency runs, and it doesn’t issue API keys for model requests. The available inspection points are your own network proxy, which sees the traffic routed through it, and the desktop’s OpenTelemetry export, which sends tool-call metadata (tool name, connector, outcome, duration, and approval status) to your collector for after-the-fact review. See Telemetry and egress for what the export can include. Chat and Cowork sessions write a local audit log to the user’s disk recording prompt text, Claude’s responses, tool inputs and results, and permission decisions. The app doesn’t send that log off the device. Server-side, the Compliance API records identity and configuration events but never tool calls or conversation content.

Data storage and retention

In Claude for Government, conversation content stays on the user’s device. If you select content categories in the Telemetry content capture setting on the Config page, Claude Desktop also sends the selected prompt, response, and tool content to your own OpenTelemetry collector, and never to Anthropic. Model requests are proxied through the Claude for Government service to the model endpoint inside the FedRAMP High boundary, and the service records only per-request metadata, not content.
  • Can Anthropic view conversations? No. Chat transcripts are stored on the user’s workstation, and the Claude for Government service does not log request or response bodies. Inference requests pass through the service to the model endpoint but are not retained.
  • Where on the device is conversation content stored? Conversation content lives under the application data directory (%LOCALAPPDATA%\Claude-3p on Windows, ~/Library/Application Support/Claude-3p on macOS, ~/.config/Claude-3p on Linux). Code session transcripts live in Claude Code’s own folder in the user’s home directory (~/.claude/projects). The application does not encrypt these files itself, so encryption at rest depends on the workstation’s full-disk encryption, such as BitLocker, FileVault, or LUKS. The operating system encryption that protects connector credentials, described under Connectors, does not apply to conversation content.
  • Is local history kept separate for each organization? No. Claude Desktop keeps conversation history in the application data directory of the operating system account in use on the device, and it does not divide that history by the Claude for Government organization or tenant the user signs in to. A user who is moved to another organization, or who signs in to a second tenant from the same operating system account, sees the same conversations and projects as before. The Claude Desktop configuration reference lists a deploymentOrganizationUuid key that separates local data by organization in other deployments. That key does not apply to Claude for Government, so leave it out of your configuration profile.
  • Projects. A project in Claude for Government is stored in the application data directory on the user’s own device, together with any instructions, links, and folder references the user adds to it. Files added to a project stay on the user’s local disk; there is no service-side project store, and files are not vectorized or indexed. Claude reads them directly from disk on demand with its file tools. A new Cowork session in the project does not reuse what an earlier session read, so Claude reads the files it needs again. The project’s memory carries short notes between sessions, such as preferences and decisions. Content Claude reads from those files is handled like the rest of the conversation: inference requests pass through the Claude for Government service to the model endpoint but are not retained.
  • Can the local data location be changed for backup or sync? The application data directory cannot be moved. It is fixed for each user, outside the Windows roaming profile, and conversations and projects do not sync between devices. To back up a user’s data or move it to a new device, copy the folders that hold it with your endpoint management tools. On the new device, the copy replaces any Claude Desktop data already there. To clear a user’s data from a device, see Removing data.
  • Can users export their data to another device? Yes, where an administrator turns on Export and import between computers, which is off by default. The user saves an unencrypted zip file of their computer account’s Claude Desktop data, and any user in the same tenant who has the setting on can import it. The file doesn’t pass through the Claude for Government service, which keeps no record of exports or imports. If you set a Telemetry endpoint, your collector gets a custom3p_device_transfer_completed log record for each saved export and finished import, with the user, the direction, and the number of sessions. It gets none while Application event level (Claude Desktop) is Off.

Web search and web fetch

Web search is operated by Anthropic inside the Claude for Government FedRAMP High boundary, and the search provider’s API is the one case where traffic egresses that boundary. Web search is built in and does not require obtaining a separate connector. Before search is enabled, an administrator must acknowledge a disclosure covering this data flow when enabling the Web search card on the Config page. By default, users approve each query before it is sent, and Claude transforms it into a generic, de-identified search request and shows the user the exact text. The Require approval for each search setting is on by default. Anthropic has a zero-data-retention agreement with the search provider. The Web search and web fetch page covers how web search is configured in other Claude Desktop deployments; the Claude for Government search path described here is specific to this deployment. In Chat and Cowork, every web page fetch is checked against your egress allowlist before the request is made, and redirects are re-checked against the allowlist on each hop. See Web search and web fetch. Allowlist entries match by name, not by the address a name resolves to; see Allowed network hosts on the Config page.

Chat and Cowork differences

Chat and Cowork share one list of projects. Projects are not shared between users. Chat conversations in a project can read the project’s memory. See Data storage and retention for where project contents are stored. Artifacts are available in both Chat and Cowork. An artifact is an interactive page that Claude writes and the Claude Desktop application displays on the user’s device, outside the sandbox, so artifacts remain available even when Advanced file analysis in Chat is off. In Chat and Cowork, Claude creates files on the user’s device in places such as these:
  • A folder the user attached to a Cowork task.
  • The working folder that each task or conversation has inside the application data directory.
  • For artifacts, the user files directory (default ~/Claude).
A user who wants results in a particular folder attaches that folder to the task and asks Claude to save the files there. Shell commands run inside the sandbox virtual machine, and on the device they can write only to the attached folders and the task’s working folder. A file that a command writes anywhere else in the virtual machine, for example under /tmp, does not appear in any folder on the device. See Chat in Claude Desktop for what a Chat conversation can reach, and User identity and local data for the folder layout.

Claude Code command-line tool

The answers below cover the standalone Claude Code command-line tool when it signs in through Claude for Government, as Connect Claude Code to Claude for Government describes. For Code sessions inside Claude Desktop, see Code in Claude Desktop. For settings that govern the tool, such as Shell command sandbox, see Claude Code on the Available settings page.
  • Which domains does Claude Code need to reach? The Claude for Government gateway address in the managed settings from Connect Claude Code to Claude for Government, over HTTPS on port 443, for sign-in, organization settings, and model inference. Sign-in finishes in the user’s browser, which needs the Claude for Government host, its sign-in service, and your agency’s identity provider, as for Claude Desktop. Claude Code also reaches, from the user’s device, the address of each connector an administrator applies to Claude Code on the Connectors card, and that connector’s sign-in service when it uses OAuth. The managed settings from Connect Claude Code to Claude for Government turn off Claude Code’s background connections before sign-in (release notes, update checks, and the official plugin marketplace).
  • Does Claude Code send telemetry to Anthropic? No. The managed settings in Connect Claude Code to Claude for Government turn off Claude Code’s usage telemetry and error reporting to Anthropic from the first launch, and the organization’s settings keep them off after sign-in.