Gemini CLI 0.36.0

What's Changed

• Changelog for v0.33.2
• feat(core): multi-registry architecture and tool filtering for subagents
• Changelog for v0.34.0-preview.4
• fix(devtools): use theme-aware text colors for console warnings and errors
• Add support for dynamic model Resolution to ModelConfigService
• chore(release): bump version to 0.36.0-nightly.20260317.2f90b4653
• fix(cli): use active sessionId in useLogger and improve resume robustness
• fix(cli): expand tilde in policy paths from settings.json
• fix(core): add actionable warnings for terminal fallbacks (#14426)
• feat(tracker): integrate task tracker protocol into core system prompt
• chore: add posttest build hooks and fix missing dependencies
• feat(a2a): add agent acknowledgment command and enhance registry discovery
• fix(cli): automatically add all VSCode workspace folders to Gemini context
• feat: add 'blocked' status to tasks and todos
• refactor(cli): remove extra newlines in ShellToolMessage.tsx
• fix(cli): lazily load settings in onModelChange to prevent stale closure data loss
• feat(core): subagent local execution and tool isolation
• fix(cli): resolve subagent grouping and UI state persistence
• refactor(ui): extract SessionBrowser search and navigation components
• fix: updates Docker image reference for GitHub MCP server
• refactor(cli): group subagent trajectory deletion and use native filesystem testing
• refactor(cli): simplify keypress and mouse providers and update tests
• Changelog for v0.34.0
• test(cli): simplify createMockSettings calls
• feat(ui): format multi-line banner warnings with a bold title
• Docs: Remove references to stale Gemini CLI file structure info
• feat(ui): remove write todo list tool from UI tips
• Fix issue where subagent thoughts are appended.
• Feat/browser privacy consent
• fix(core): explicitly map execution context in LocalAgentExecutor
• feat(plan): support plan mode in non-interactive mode
• feat(core): implement strict macOS sandboxing using Seatbelt allowlist
• docs: add additional notes
• fix(cli): resolve duplicate footer on tool cancel via ESC (#21743)
• Changelog for v0.35.0-preview.1
• fix(ui): fix flickering on small terminal heights
• fix(acp): provide more meta in tool_call_update
• docs: add FAQ entry for checking Gemini CLI version
• feat(core): resilient subagent tool rejection with contextual feedback
• fix(cli): correctly handle auto-update for standalone binaries
• feat(core): add content-utils
• fix: circumvent genai sdk requirement for api key when using gateway auth via ACP
• fix(core): don't persist browser consent sentinel in non-interactive mode
• fix(core): narrow browser agent description to prevent stealing URL tasks from web_fetch
• feat(cli): Partial threading of AgentLoopContext.
• fix(browser-agent): enable "Allow all server tools" session policy
• refactor(cli): integrate real config loading into async test utils
• feat(core): inject memory and JIT context into subagents
• Fix logging and virtual list.
• feat(core): cap JIT context upward traversal at git root
• Docs: Minor style updates from initial docs audit.
• feat(core): add experimental memory manager agent to replace save_memory tool
• Changelog for v0.35.0-preview.2
• Update website issue template for label and title
• fix: upgrade ACP SDK from 0.12 to 0.16.1
• Update callouts to work on github.
• feat: ACP: Add token usage metadata to the send method's return value
• fix(plan): clarify that plan mode policies are combined with normal mode
• Add ModelChain support to ModelConfigService and make ModelDialog dynamic
• Ensure that copied extensions are writable in the user's local directory
• feat(core): implement native Windows sandboxing
• feat(core): add support for admin-forced MCP server installations
• chore(lint): ignore.gemini directory and recursive node_modules
• feat(cli): conditionally exclude ask_user tool in ACP mode
• feat(core): introduce AgentSession and rename stream events to agent events
• feat(worktree): add Git worktree support for isolated parallel sessions
• Add support for linking in the extension registry
• feat(extensions): add --skip-settings flag to install command
• feat(telemetry): track if session is running in a Git worktree
• refactor(core): use absolute paths in GEMINI.md context markers
• fix(core): add sanitization to sub agent thoughts and centralize utilities
• feat(core): refine User-Agent for VS Code traffic (unified format)
• Fix schema for ModelChains
• test(cli): refactor tests for async render utilities
• feat(core): add security prompt for browser agent
• refactor(ide): replace dynamic undici import with static fetch import
• test(cli): address unresolved feedback from PR #23252
• feat(browser): add sensitive action controls and read-only noise reduction
• Disabling failing test while investigating
• fix broken extension link in hooks guide
• fix(core): fix agent description indentation
• Wrap the text under TOML rule for easier readability in policy-engine.md…
• fix(extensions): revert broken extension removal behavior
• feat(core): set up onboarding telemetry
• Retry evals on API error.
• fix(evals): remove tool restrictions and add compile-time guards
• fix(hooks): support 'ask' decision for BeforeTool hooks
• feat(browser): add warning message for session mode 'existing'
• chore(lint): enforce zero warnings and cleanup syntax restrictions
• fix(cli): add Esc instruction to HooksDialog footer
• Disallow and suppress misused spread operator.
• fix(core): refine CliHelpAgent description for better delegation
• fix(core): enable global session and persistent approval for web_fetch
• fix(plan): add state transition override to prevent plan mode freeze
• fix(cli): record skill activation tool calls in chat history
• fix(core): ensure subagent tool updates apply configuration overrides immediately
• fix(cli): resolve flicker at boundaries of list in BaseSelectionList
• test(cli): force generic terminal in tests to fix snapshot failures
• Evals: PR Guidance adding workflow
• feat(core): refactor SandboxManager to a stateless architecture and introduce explicit Deny interface
• feat(core): add event-translator and update agent types
• perf(cli): parallelize and background startup cleanup tasks
• fix: "allow always" for commands with paths
• fix(cli): prevent terminal escape sequences from leaking on exit
• feat(cli): implement full "GEMINI CLI" logo for logged-out state
• fix(plan): reserve minimum height for selection list in AskUserDialog
• fix(core): harden AgentSession replay semantics
• test(core): migrate hook tests to scheduler
• chore(config): disable agents by default
• fix(ui): make tool confirmations take up entire terminal height
• fix(core): prevent redundant remote agent loading on model switch
• refactor(core): update production type imports from coreToolScheduler
• feat(cli): always prefix extension skills with colon separator
• fix(core): properly support allowRedirect in policy engine
• fix(cli): prevent subcommand shadowing and skip auth for commands
• fix(test): move flaky tests to non-blocking suite
• Changelog for v0.35.0-preview.3
• feat(skills): add behavioral-evals skill with fixing and promoting guides
• refactor(core): delete obsolete coreToolScheduler
• Changelog for v0.35.0-preview.4
• feat(core): add LegacyAgentSession
• feat(test-utils): add TestMcpServerBuilder and support in TestRig
• fix(core)!: Force policy config to specify toolName
• eval(save_memory): add multi-turn interactive evals for memoryManager
• fix(telemetry): patch memory leak and enforce logPrompts privacy
• perf(cli): background IDE client to speed up initialization
• fix(cli): prevent Ctrl+D exit when input buffer is not empty
• fix: ACP: separate conversational text from execute tool command title
• feat(evals): add behavioral evaluations for subagent routing
• refactor(cli,core): foundational layout, identity management, and type safety
• fix(core): accurately reflect subagent tool failure in UI
• Changelog for v0.35.0-preview.5
• feat(ui): implement refreshed UX for Composer layout
• fix: API key input dialog user interaction when selected Gemini API Key
• docs: update /mcp refresh to /mcp reload
• Implementation of sandbox "Write-Protected" Governance Files
• feat(sandbox): dynamic macOS sandbox expansion and worktree support
• fix(acp): Pass the cwd to AcpFileSystemService to avoid looping failures in asking for perms to write plan md file
• fix(plan): sandbox path resolution in Plan Mode to prevent hallucinations
• feat(ui): allow immediate user input during startup
• refactor(sandbox): reorganize Windows sandbox files
• fix(core): improve remote agent streaming UI and UX
• perf(cli): optimize --version startup time
• refactor(core): stop gemini CLI from producing unsafe casts
• use enableAutoUpdate in test rig
• feat(core): change user-facing auth type from oauth2 to oauth
• chore(deps): fix npm audit vulnerabilities
• test(evals): fix overlapping act() deadlock in app-test-helper
• fix(patch): cherry-pick 055ff92 to release/v0.36.0-preview.0-pr-23672 to patch version v0.36.0-preview.0 and create version 0.36.0-preview.1
• fix(patch): cherry-pick 765fb67 to release/v0.36.0-preview.5-pr-24055 to patch version v0.36.0-preview.5 and create version 0.36.0-preview.6