Summary
Zed 1.14.2 adds sandboxing for the built-in Agent's terminal and fetch tools, isolating the shell commands and web requests the agent runs. It tightens the boundary around autonomous actions taken by Zed's AI agent inside the editor.
What changed
On August 5, 2026 Zed released 1.14.2, which added sandboxing for the Agent's terminal and fetch tools so the commands and web fetches the agent issues run inside a constrained environment rather than with the editor's full access.
Why it matters
As editors let AI agents run shell commands and fetch from the web autonomously, those tool calls become a real attack and accident surface. Sandboxing the terminal and fetch tools reduces the blast radius of a misbehaving or prompt-injected agent, part of a broader move to make agent execution safe by default rather than fully trusted.
Evidence excerpt
Added sandboxing for the Agent's terminal and fetch tools.