Gemini CLI 0.60 hardens extensions, sandboxes, and MCP OAuth
Gemini CLI 0.60.0 adds extension consent and isolation, stronger workspace and sandbox boundaries, safer web fetch routing, and strict MCP OAuth issuer checks.
Editorial Team
The Authority AI Tools editorial team maintains this directory using vendor documentation, dated source checks, product changelogs, and clearly identified hands-on observations where available.
Google’s open-source Gemini CLI shipped stable v0.60.0 on September 15, 2026. This release is less about a new model than about tightening the boundaries around extensions, sandboxes, web fetching, and Model Context Protocol (MCP) authentication.
Extensions now ask before changing the environment
Extensions can request environment changes, but Gemini CLI now prompts for user consent and sanitizes runtime-altering environment variables. The goal is to make extension behavior more explicit instead of allowing an installed extension to silently alter the execution environment.
The release also strengthens extension path resolution and workspace-boundary validation. Symlink handling is hardened, and the CLI mitigates NTFS 8.3 short-name paths in command-safety and file-discovery utilities. These controls matter when an agent can inspect or modify files across a large or shared workspace.
Stronger sandbox and output boundaries
Version 0.60.0 isolates settings and temporary directories inside sandbox containers and adds macOS Seatbelt path isolation. It also tightens permission and ownership checks on system-wide configuration paths.
The core now enforces envelope metadata provenance for untrusted tool outputs. That is useful for reducing ambiguity about where tool results came from, but it does not make an external tool trustworthy by itself. Teams should continue to review MCP servers, extensions, and shell permissions before enabling them in sensitive repositories.
Open-source, model-agnostic AI coding assistant for VS Code and JetBrains
MCP OAuth and web fetch safety
The release implements strict RFC 9207 issuer identification in the MCP OAuth flow. It also improves destination validation and connection routing in web-fetch utilities. Together, these changes reduce the chance that a connection or authentication flow is accepted merely because it looks structurally valid.
The stable release is marked latest by the Gemini CLI project. Upgrade testing should include custom extensions, MCP OAuth login, sandboxed commands, symlink-heavy repositories, and any web-fetch workflow that relies on redirects or private-network blocking.
Sources
- Gemini CLI — stable v0.60.0 changelog (September 15, 2026): https://geminicli.com/docs/changelogs/latest/
- Gemini CLI — official v0.60.0 release: https://github.com/google-gemini/gemini-cli/releases/tag/v0.60.0
- Gemini CLI — complete changelogs: https://geminicli.com/docs/changelogs/
- Google Gemini CLI — official repository: https://github.com/google-gemini/gemini-cli
- Google for Developers on X — official developer updates: https://x.com/googledevs
Tools Mentioned in This Article
Free Resource
2026 AI Coding Tools Comparison Chart
Side-by-side comparison of features, pricing, and capabilities for every major AI coding tool.
No spam, unsubscribe anytime.
Workflow Resources
Frequently Asked Questions
What is Gemini CLI 0.60 hardens extensions, sandboxes, and MCP OAuth?
Related Articles
Cursor discloses malicious Claude Code extension in its VS Code marketplace
Cursor says a modified Anthropic.claude-code 2.1.313 extension was available for part of October 29; potentially affected users should follow Cursor's official response.
Read more →Product UpdatesOpenAI adds monthly API credits for eligible Pro and Business plans
Eligible ChatGPT Pro and Business subscribers can claim promotional API credits, with plan-specific amounts and a $500 monthly cap for Business workspaces.
Read more →Product UpdatesGoogle releases the Gemini API CLI in beta for agents and CI
The stateless Gemini API CLI lets coding agents, scripts, and CI pipelines use the Interactions API for generation, file analysis, and managed agents.
Read more →