GitHub Copilot for JetBrains IDEs has added a suite of public‑preview features, most notably enterprise‑managed sandbox policies, cross‑file cursor jumps in edit suggestions, the ability to add global files to chat context, diagnostics for policy enforcement, and a new /ide command that ties Copilot CLI sessions to the IDE. These updates give administrators direct control over sandbox behavior and give developers smoother multi‑file interactions, which directly affect how you design, operate, and secure AI‑assisted development pipelines.
Enterprise‑Managed Sandbox Policies
Administrators can now define sandbox settings centrally for all JetBrains users. The managed policies cover enablement, filesystem and network access, proxy configuration, developer‑tool access, and macOS Keychain usage. When a policy is active, the corresponding UI controls in the IDE are locked, and the IDE surface indicates which settings are governed by the organization. The sandbox UI appears only if the organization enables the Editor Preview flag or explicitly configures a managed setting to toggle the sandbox.
Cross‑File Cursor Jumps in Next Edit Suggestions
Next‑edit suggestions can now move the cursor to a different file when a change spans multiple files. Instead of manually locating the continuation point, the IDE jumps directly to the relevant location, streamlining coordinated refactors and reducing context‑switch overhead.
Global Project Context in Chat
Users can now attach entire directories or files to the Copilot chat context. This reduces repetitive file‑by‑file additions and helps the model generate responses that consider project‑wide information, which is especially useful for changes that affect many components.
Enterprise Policy Diagnostics
A diagnostics view lets users verify that the sandbox and other policies are correctly detected on their machine. This provides a quick sanity check that the local configuration complies with organizational requirements.
Connecting Terminal Copilot CLI Sessions to the IDE
The new /ide command links a terminal Copilot CLI session with the current JetBrains context, including selections, diagnostics, and file references. Environment variables from the IDE terminal and the project’s Python interpreter are inherited, allowing the CLI to activate the project’s virtual environment and execute commands that match the IDE’s runtime.
Additional UX and Reliability Improvements
Other refinements include a larger ask‑user card for long prompts, a streamlined model picker with BYOK grouping, copy‑support for agent debug logs, and better handling of MCP configuration windows. Reliability across MCP servers and agent sessions has been hardened, addressing issues such as blank chat sessions, incorrect working‑set counts, and read‑only file editing problems.
Related CloudNinjas coverage: AI engineering.
What This Means For Practitioners
Architectural impact: Centralized sandbox policies enable a consistent execution boundary across developer machines, reducing variance in what code can access during AI‑generated suggestions. Teams should review existing sandbox configurations and align them with the new managed settings to avoid unexpected lock‑outs.
Implementation considerations: Enable the Editor Preview flag or configure managed sandbox settings to surface the new UI controls. Incorporate the /ide workflow into scripts that rely on Copilot CLI to keep terminal actions in sync with the IDE state.
Operational actions: Use the policy diagnostics view to audit compliance on each workstation. Update CI/CD pipelines that invoke Copilot CLI to ensure they inherit the correct environment variables and virtual environments.
Security perspective: The ability to restrict filesystem, network, and Keychain access at the policy level provides a clearer security perimeter for AI‑assisted code generation. Verify that the enforced policies match your organization’s threat model, especially for projects handling sensitive data.
Going forward, monitor the public‑preview status of these features and plan for a migration path to fully managed sandbox enforcement. Evaluate whether the cross‑file cursor jumps and global chat context reduce manual steps in your typical refactor cycles, and adjust tooling or training accordingly.
