Gemini CLI v0.37.0: The experimental browser agent, security and user experience enhancements

[srithreepo]

The v0.37.0 release is focused on experimental browser agent, security and user experience enhancements.

🧪 Experimental

• 🌐 Browser Agent – Gemini CLI Can now connect and interact with Chrome.
  The experimental Browser Agent is now available on Gemini CLI, and we’d love for you to put it to the test! Gemini CLI can use the Browser Agent to navigate, interact with, and automate complex tasks across the live web—all while keeping your main session context clean and focused. [docs]
• 🔩 Few things for you to try...
  • Enable the agent by adding "browser_agent": { "enabled": true } to your agents.overrides in settings.json. [docs]
  • Use the @browser_agent command to trigger it for specific tasks, like filling out forms, navigating complex UI, or extracting data.
  • Try natural language requests like "Go to the Gemini CLI release page and summarize the latest changes" to see it in action.

🛡️ Security & File Integrity for Sandboxing – We’re giving you more control over your workflows when using Sandbox mode. [docs]

• The new forbiddenPaths feature allows the sandbox to automatically respect your project's ignore files (like .gitignore), ensuring the model never touches restricted areas. (feat(sandbox): implement forbiddenPaths for OS-specific sandbox managers #23282, @ehedlund)
• Additionally, we’ve introduced a dedicated lockdown for environment files to keep your secrets and API keys strictly private. (feat(sandbox): implement secret visibility lockdown for env files #23712, @DavidAPierce)

✨ Refreshed UI & Navigation

• Tokyo Night Theme: A beautiful new dark theme option is now available. (feat(ui): add Tokyo Night theme #24054, @danrneal)
• Accelerated Scrolling: Improved navigation with accelerated scrolling performance in alternate buffer modes. (fix(ui): add accelerated scrolling on alternate buffer mode #23940, @devr0306 )
• Clearer Status: The footer now explicitly displays your current Authentication status and provides more compact tool output layouts. (feat(cli): add auth info to footer #24042, @sehoon38 )
• Immediate Queueing: You can now use the 'Tab' key to queue up messages while the model is still generating a response. (feat(ui): allow immediate user input during startup #23661, @sehoon38)

🧭 Platform Support

• Windows & Linux Expansion: Added support for dynamic sandbox expansion and Git Worktree support for both Linux and Windows. (feat(sandbox): dynamic Linux sandbox expansion and worktree support #23692, @galz10 )
• Model Routing: Improved dynamic routing for Gemini 3.1 Pro and enabled mid-stream retries for better resilience. (Fix dynamic model routing for gemini 3.1 pro to customtools model #23641, @kevinjwang1 )

🎉 Gemini CLI Extensions

• Google Cloud (gcloud): Enable MCP-compatible AI agents to interact with your Google Cloud environment using natural language. Automate complex workflows and manage cloud resources without memorizing complex command syntax.
  gemini extensions install https://github.com/gemini-cli-extensions/gcloud

Small features, Polish, Reliability & Bug Fixes: A large amount of changes, smaller features, UI updates, reliability and bug fixes + general polish made it in this week!

[pala20cm]

sex

[wissamblue69-dotcom]

The v0.37.0 release is focused on experimental browser agent, security and user experience enhancements.

🧪 Experimental

• 🌐 Browser Agent – Gemini CLI Can now connect and interact with Chrome.
  The experimental Browser Agent is now available on Gemini CLI, and we’d love for you to put it to the test! Gemini CLI can use the Browser Agent to navigate, interact with, and automate complex tasks across the live web—all while keeping your main session context clean and focused. [docs]
• 🔩 Few things for you to try...
  • Enable the agent by adding "browser_agent": { "enabled": true } to your agents.overrides in settings.json. [docs]
  • Use the @browser_agent command to trigger it for specific tasks, like filling out forms, navigating complex UI, or extracting data.
  • Try natural language requests like "Go to the Gemini CLI release page and summarize the latest changes" to see it in action.

🛡️ Security & File Integrity for Sandboxing – We’re giving you more control over your workflows when using Sandbox mode. [docs]

• The new forbiddenPaths feature allows the sandbox to automatically respect your project's ignore files (like .gitignore), ensuring the model never touches restricted areas. (feat(sandbox): implement forbiddenPaths for OS-specific sandbox managers #23282, @ehedlund)
• Additionally, we’ve introduced a dedicated lockdown for environment files to keep your secrets and API keys strictly private. (feat(sandbox): implement secret visibility lockdown for env files #23712, @DavidAPierce)

✨ Refreshed UI & Navigation

• Tokyo Night Theme: A beautiful new dark theme option is now available. (feat(ui): add Tokyo Night theme #24054, @danrneal)
• Accelerated Scrolling: Improved navigation with accelerated scrolling performance in alternate buffer modes. (fix(ui): add accelerated scrolling on alternate buffer mode #23940, @devr0306 )
• Clearer Status: The footer now explicitly displays your current Authentication status and provides more compact tool output layouts. (feat(cli): add auth info to footer #24042, @sehoon38 )
• Immediate Queueing: You can now use the 'Tab' key to queue up messages while the model is still generating a response. (feat(ui): allow immediate user input during startup #23661, @sehoon38)

🧭 Platform Support

• Windows & Linux Expansion: Added support for dynamic sandbox expansion and Git Worktree support for both Linux and Windows. (feat(sandbox): dynamic Linux sandbox expansion and worktree support #23692, @galz10 )
• Model Routing: Improved dynamic routing for Gemini 3.1 Pro and enabled mid-stream retries for better resilience. (Fix dynamic model routing for gemini 3.1 pro to customtools model #23641, @kevinjwang1 )

🎉 Gemini CLI Extensions

• Google Cloud (gcloud): Enable MCP-compatible AI agents to interact with your Google Cloud environment using natural language. Automate complex workflows and manage cloud resources without memorizing complex command syntax.
  gemini extensions install https://github.com/gemini-cli-extensions/gcloud

Small features, Polish, Reliability & Bug Fixes: A large amount of changes, smaller features, UI updates, reliability and bug fixes + general polish made it in this week!