Zed Editor v1.14.1-pre リリースZed Editor Releases v1.14.1-pre

Zed, the Rust-based code editor known for its speed and its expanding set of AI features, has published v1.14.1-pre, a release">pre-release build that concentrates on making its built-in Agent safer to run while smoothing several everyday editing and version-control workflows. The headline change is sandboxing for the Agent's terminal commands and web fetches, a move that reflects a broader industry conversation about how much latitude AI coding assistants should have on a developer's machine.

Sandboxing, in this context, refers to constraining the operations an AI agent performs so that they run in a more isolated environment rather than with unrestricted access to the host system. Because agentic coding tools increasingly execute shell commands, install dependencies, and reach out to the network on a user's behalf, the risk surface grows accordingly. Applying sandboxing to both terminal commands and web fetches appears intended to limit the blast radius if an agent runs something unexpected or acts on untrusted input. The release notes frame this as a security-oriented improvement, though the exact boundaries of the sandbox and the platforms it covers are best confirmed against Zed's own documentation.

The build also adds undo and redo support for file operations in the Project Panel, with accompanying documentation. The Project Panel is Zed's file-tree sidebar, and operations such as creating, renaming, moving, or deleting files have historically been difficult to reverse in many editors once performed. Bringing these actions under the same undo and redo model that developers expect from text editing is a usability refinement that reduces the cost of mistakes during routine file management.

On the version-control side, the release introduces a Skip Hooks option for commits. Git hooks are scripts that run automatically at certain points in the version-control lifecycle, such as pre-commit hooks that lint code, run tests, or enforce formatting before a commit is accepted. These checks are valuable but can be slow or occasionally get in the way of a quick, intentional commit. A Skip Hooks option gives developers a deliberate way to bypass those scripts when needed, mirroring the command-line behavior of committing with hooks disabled, without leaving the editor.

Rounding out the changes, the update makes Agent Panel fonts configurable. The Agent Panel is the interface through which Zed's AI assistant is used, and allowing users to set its font is a smaller quality-of-life adjustment that brings the panel in line with the customization already available elsewhere in the editor. Collectively, these four items suggest a release cycle balancing security hardening with incremental interface polish rather than a single large feature push.

It is worth underscoring that this is a release">pre-release, denoted by the "-pre" suffix on the version number. release">Pre-release and nightly channels typically ship newer changes to users who want early access and are willing to tolerate rough edges, ahead of a more thoroughly tested stable build. Features that land in a release">pre-release can shift before they reach the general release, so behavior described here may still evolve. The release was, according to the notes, shipped by the Zed Guild, the team's framing for the group responsible for a given cycle.

For context, Zed was created by the team behind the Atom editor and the Tree-sitter parsing library, and it has positioned itself as a high-performance, collaborative editor written in Rust. Its investment in an integrated Agent places it alongside a wave of AI-forward development tools, including Cursor, GitHub Copilot's agent modes, and command-line assistants such as Claude Code, all of which are grappling with the same tension between capability and safety. As these tools take on more autonomous actions, sandboxing and permission controls are becoming a recurring theme across the category rather than a differentiator unique to any one product.

For users tracking Zed's trajectory, v1.14.1-pre is likely most interesting for what the sandboxing work signals about the editor's direction: treating the Agent as a component that needs guardrails by default. Developers who rely on custom Git hooks, frequent file reorganization, or a heavily used Agent Panel stand to benefit most immediately from the remaining changes. As always with a release">pre-release, consulting the linked documentation and testing in a non-critical project is the prudent approach before depending on the new behavior in day-to-day work.