Zum Hauptinhalt springen

Antigravity

Changelog

Website besuchen

Googles KI-gestützte Agentic IDE

Erfasste Versionshinweise aus offiziellen Quellen. Ältere Releases können fehlen.

Versionshinweise ohne deutsche Übersetzung stehen im englischen Original.

Neueste Version:v1.2.9
Erfasste Releases:136
Quelle:Web Changelog
RSS

Releases (136)

v1.2.9

Antigravity CLI: Direct subagent messaging syntax, Vim numeric count multipliers, headless lifecycle handling, and concurrency fixes

23. September 2026

  • •Added @ prompt syntax to send a message directly to a subagent conversation, with autocomplete listing running and completed subagents.
  • •Added Vim numeric count multipliers so counts apply to operators, motions, and actions in Normal and Visual modes, including 3dw, 2d3w, 3dd, 3x, 3rX, 3p, 3u, [count]G/gg/$, and counted text objects such as 2di(.
  • •Improved GEMINI_API_KEY sessions to reduce behavior discrepancies with the non-API-key sign-in path.
  • •Improved the artifact viewer to show a Left/Right pan hint in the footer when a wide diagram overflows the window.
  • •Improved /rewind to show a relative timestamp for each step and to focus the most recent step when the panel opens.
  • •Improved command allow-listing suggestions to recognize jj config and jj op subcommands when offering to always allow a command.
  • •Fixed headless (-p / --prompt) runs leaving daemon background processes running after exit, which could hang scripts reading the CLI's output until end-of-file; daemon processes now terminate when the run ends.
  • •Fixed headless (-p / --prompt) runs cancelling still-running background tasks about 5 seconds after the agent went idle; runs now wait for background tasks until the --print-timeout deadline, up to a 30-minute cap.
  • •Fixed a conversation-history database corruption risk where checking the database file for write access could silently drop file locks held by concurrent CLI processes on the same file.
  • •Fixed context compaction failing when the tool configuration used for compaction checkpoints was rejected in certain scenarios.
  • •Fixed a backend crash when a streamed model response chunk arrived without its response envelope, which terminated the session with connection errors.
  • •Fixed markdown table column alignment when a table cell contains file links that wrap across lines.
  • •Fixed markdown file links to code symbols dropping their display text and rendering the raw path instead.
  • •Fixed incomplete enterprise sign-ins (for example closing the browser window before finishing license or project selection) leaving behind a stuck partial credential; such credentials are now cleared automatically so sign-in can be retried cleanly.
  • •Fixed the browser companion page title to read Antigravity CLI instead of Antigravity Cli.
v2.17.0

Plan before you build

22. September 2026

  • •Reveal in Finder for projects. Projects in the sidebar have a new Reveal in Finder entry (Show in File Explorer on Windows) in their overflow menu, opening the project folder in your system file manager. Multi-folder projects open one window per folder.
  • •Toggle theme from the command palette. A new "Toggle Between Light/Dark Themes" command flips the theme without a trip to Settings. On System, the first toggle pins the opposite of whatever your OS resolved to.
  • •Custom agents can declare their own hooks. A Markdown agent can list hook files in its front matter with a hooks: entry. Paths may be relative to the agent, relative to the project, or absolute.
  • •Save Video. Right-click a video anywhere it appears — in the chat, the file viewer, a browser recording, or the full-screen preview — and choose Save Video.
  • •Rules no longer crowd out everything else. Rules get their own 20,000-token budget, so a large rules file stops pushing your skills, subagents and MCP tools out of the agent's context. A rule over budget is listed by file path and description for the agent to open when relevant, instead of being cut short.
  • •Per-project configuration moved. Repository customization settings now load from /.gemini/config.json. The older .agents/settings.json is no longer read — move any personal_customization_dir entry into the new file.
  • •Paste properly into the chat box. Right-clicking the chat input and choosing Paste handles images and rich text, keeping links as markdown, instead of only plain text. Pasting over a selected attachment no longer drops the pasted text.
  • •Slash commands and plugins read like products. The pill inserted when you pick a slash command shows the skill's proper name and logo instead of its raw command name, and the built-in Google plugins show readable names — Chrome DevTools, Google Maps Platform, Gemini API, Android CLI, Data Agent Kit, Modern Web Guidance, Google Antigravity SDK.
  • •Fewer link cards you did not ask for. The agent surfaces a link card in the conversation only when there is something to open or act on, such as a review it created, a document it edited, or a local server it started. Links it merely consulted are bookmarked to the sidebar quietly.
  • •Sidebar polish. Hovering a conversation highlights the whole row rather than just the text, and a sidebar that closed itself because you narrowed the window reopens when you widen it again. A sidebar you closed yourself stays closed.
  • •.cppm files are syntax-highlighted as C++ and get the code file icon.
  • •Rewinding a conversation no longer restores out-of-date files. The snapshot search used to sweep backward from the point you rewound to, so it could hand back a state from earlier in the conversation. It now takes the first snapshot at or after that point and never an earlier one, and replays the steps when no such snapshot exists.
  • •Large files and diffs no longer freeze the app. Syntax highlighting and line diffing run off the main thread, and files over 100 KB fall back to plain text.
  • •Custom rules, workflows and agents load on large projects. They no longer report a load failure when scanning your project folders takes more than three seconds; the timeout is now fifteen seconds, matching what skills already had.
  • •/btw keeps your agent. A side question runs with the same agent and settings as the conversation you asked it in, instead of falling back to the default agent.
v2.16.0

Windows Subsystem for Linux, Office attachments, and live subagent cards

22. September 2026

  • •Added a Windows Subsystem for Linux (WSL) section to Application Settings on Windows, so you can connect to an installed WSL distribution or switch back to your local Windows environment. When connected, the title bar shows the active distribution name.
  • •You can now attach Microsoft Office documents (Word, Excel, and PowerPoint) to a prompt by dragging and dropping them into the message box, and Gemini models read them natively instead of falling back to opening the stored file.
  • •Subagent invocations now appear as dedicated cards in the conversation stream with live running, waiting, and completed status, a hover control to stop a running subagent, and one-click navigation to its conversation in the side pane.
  • •Prompt actions in interactive widgets and HTML artifacts now fill in and focus the message box instead of sending automatically, so you can review or edit the prompt before submitting it.
  • •Right-click an image in an attachment, a preview modal, or an earlier message to Copy Image or Save Image.
  • •SVG images can now be copied to the clipboard and pasted into other applications.
  • •The loading indicator now shows what the agent is doing while it prepares a response, including tool initialization progress and a live countdown with attempt count when a model request is retried.
  • •The timestamp in the agent's message footer now shows when the response finished rather than when it started, matching the duration shown on long-running turns.
  • •Application keyboard shortcuts now take priority over web previews, so shortcuts keep working while a preview has focus.
  • •The Confirm Undo dialog now keeps the file list at a fixed size and scrolls it in place, so Confirm and Cancel stay visible when reverting many files.
  • •Inline pills and icons inside markdown headings now scale with the heading's size and weight instead of rendering small.
  • •Discovering skills and other customizations is now dramatically faster on large projects and network-mounted folders, cutting a startup scan that previously took over a minute to under a second.
  • •Fixed an issue where browser sign-in could fail permanently if the app disconnected, reloaded, or the computer went to sleep while you were completing sign-in in your browser.
  • •Fixed an issue where expired or rejected enterprise sign-in credentials stayed cached, leaving the app in a state you could not sign back in from.
  • •Fixed an issue where web search returned no summary when the search summarizer used a reasoning model.
v1.2.8

Antigravity CLI: Proportional context compaction budgets, full-window sizing, custom model media support, and leak fixes

22. September 2026

  • •Improved context compaction to spread its user-request budget across all captured prompts, so a long initial instruction is no longer truncated to a small fixed slice when the other prompts in the conversation are short.
  • •Improved context compaction to size summary and truncation budgets from the model's full context window instead of the compaction trigger threshold, so compacted summaries and background-task lists are no longer prematurely cut short.
  • •Fixed PDF and audio tool outputs and attachments failing with unsupported mime type errors on custom models configured in settings.json; custom models now accept PDFs by default and honor the modelFeatures media flags for images, video, PDF, and audio.
  • •Fixed pressing Ctrl+G to edit the prompt in a full-screen terminal editor such as vim hanging with Vim: Warning: Output is not to a terminal; external editors now run against a real terminal stdout.
  • •Fixed voice dictation sessions longer than 4 minutes failing with a deadline error and erasing the drafted transcription; recordings now automatically stop and finalize at 3 minutes 30 seconds.
  • •Fixed the startup banner displaying a Google Cloud Project ID for consumer (non-enterprise) sign-ins.
  • •Fixed a stack-overflow crash when loading or compacting conversations containing background-task, subagent-management, messaging, or scheduling steps.
  • •Fixed forked conversations inheriting step output data from steps after the fork point, which could collide with new steps produced in the fork.
  • •Fixed a background poller and timer leaking for every conversation session, slowly accumulating memory over long-running sessions.
  • •Fixed canceled conversation and workspace creation requests continuing to provision resources in the background; aborted requests now stop immediately.
  • •Fixed quitting the CLI taking around 5 extra seconds before the process exited; shutdown now cancels open streaming connections immediately instead of waiting for a forced timeout.
v0.1.18

Antigravity IDE: Local model support with LiteRT and LocalOpenAI, standardized evaluation presets, and custom subagent models

21. September 2026

  • •Antigravity SDK + Local Models Support: Official local model support is now ready to use with first-class LiteRTAgentConfig and LocalOpenAIAgentConfig configurations. Developers can execute on-device models with automated lightweight presets or integrate with local OpenAI-compatible endpoints.
  • •Standardized Evaluation Preset (AgentConfig.eval()): Adds a standardized, benchmark-ready preset that configures autonomous permissions, benchmark retry behavior, daemon execution in commands, and strips image generation and subagents to focus on core coding evaluations.
  • •Custom Subagent Models: Allows developers to assign specific model targets to subagents independently of the root agent configuration.
  • •Built-in Schedule Tool: Enables the schedule tool by default across standard tool groups, permitting agents and subagents to set timers and cron-based background jobs alongside task management.
  • •Automatic Lightweight Presets for LiteRT: Instantiating LiteRTAgentConfig now automatically applies optimized lightweight presets—including reduced prompt overhead and synchronous context compaction—without requiring an explicit call to .lightweight().
  • •Task Management Pairing: Automatically enables the manage_task tool whenever run_command or schedule is active, allowing background task lifecycle management.
  • •LiteRT & Local Model Examples: Added getting-started guides and end-to-end examples demonstrating on-device execution with LiteRTAgentConfig (Gemma 4 26B) and OpenAI-compatible local endpoints via LocalOpenAIAgentConfig.
  • •Sandbox Availability Warning: Added a session startup warning when enable_sandbox=True is requested on an environment or OS backend where sandbox isolation cannot be enforced.
  • •Extended JSON Schema Normalization: Added schema normalization support for OpenAPI and JSON Schema Draft 7 / 2020-12 keywords (such as multipleOf, prefixItems, and dependentSchemas) and prevented accidental mutation of uppercase sample values.
  • •Optimized ToolRunner Coercion: Improved type resolution for closure-scoped tools and forward-referenced ToolContext parameters, and introduced TypeAdapter caching to accelerate tool call execution.
  • •Top-Level ServiceTier Export: Re-exported ServiceTier at the root package namespace for easier import parity.
  • •Excluded ASK_QUESTION from Default Tools: Excluded BuiltinTools.ASK_QUESTION from BuiltinTools.default(). Default configurations run autonomously; agents in headless workflows will no longer attempt interactive user prompts. To re-enable interactive questions, explicitly pass BuiltinTools.ASK_QUESTION in enabled_tools.
  • •Single Compaction Threshold Dial: Simplified CompactionConfig to a single token_threshold property, deprecating legacy context token limits and interval dials. To configure context compaction, specify CompactionConfig(token_threshold=...).
  • •Service Tier Ingestion: Fixed an unhandled ValueError when connecting via gateways reporting unlisted backend service tiers (such as Vertex AI PROVISIONED_THROUGHPUT) by safely ignoring unknown tier values while preserving token counts.
  • •Removed Deprecated modified_arguments_json: Removed the legacy JSON string fallback in tool hook interception in favor of modified_args.
v2.15.1

Windows sandbox support

19. September 2026

  • •File and network sandboxing on Windows are now supported.
v1.2.7

Antigravity CLI: Inline Kitty graphics for math and Mermaid diagrams, interactive ask_question navigation, dedicated rule token budgets, and 30s retry backoff

19. September 2026

  • •Added inline Kitty graphics rendering for LaTeX math equations ($$...$$ and fenced math, latex, and tex blocks) and Mermaid flowcharts and sequence diagrams in the artifact detail viewer, with Unicode ASCII fallback, Left/Right arrow horizontal panning for wide diagrams, and m to cycle between Image, ASCII, and Raw views.
  • •Improved the ask_question interactive prompt with Left/Right arrow navigation between questions, inline previews of confirmed write-in answers, check marks on answered options, and single-step submission on the final question.
  • •Improved the CLI startup banner for enterprise accounts to display the active Google Cloud Project ID beneath the signed-in account and plan tier.
  • •Improved model API retry responsiveness by capping per-attempt retry backoff at 30 seconds instead of waiting up to 4 minutes between attempts.
  • •Improved customization token budgeting by giving user and workspace rules a dedicated 20,000-token budget (cutting oversized rules on newline boundaries and listing over-budget rules by path and description) so large rule sets no longer evict skills, workflows, subagents, or MCP tools.
  • •Improved the /usage panel by removing the duplicate remaining-quota percentage line beneath each progress bar.
  • •Improved the default agent and subagent toolset by retiring the legacy find_by_name, grep_search, and list_dir tools from the default baseline while keeping them available to custom agents that explicitly list them in tools.
  • •Fixed plugin skill slash commands being double-prefixed (/::) when a skill's frontmatter name already includes the plugin prefix, or being shadowed when multiple plugins define skills with the same short name.
  • •Fixed plugin upgrades leaving old MCP server and sidecar processes running from deleted install directories, and pruned superseded plugin versions and staging directories from the marketplace cache on startup.
  • •Fixed disabled plugins disappearing from the /plugin list after enabled-only customization queries, and hardened plugin ID path validation on uninstall.
  • •Fixed terminal rendering and Kitty keyboard protocol stack handling on exit and screen clear by upgrading Bubble Tea to v2.0.9.
  • •Fixed headless (-p / --prompt) runs occasionally skipping the background-task waiting notice, logged background SDK tool progress to task log files, and reduced memory usage by cloning truncated command output previews.
  • •Fixed starting the CLI unpinning conversations or marking them unread in the desktop app.
v2.15.0

Custom agent controls and keyboard navigation

18. September 2026

  • •Custom agents can now switch off the default prompt sections and default tools and add back only the tools they need.
  • •Page Up, Page Down, Home, and End now scroll through a conversation, and Escape steps out of the prompt box.
  • •The custom agent you pick in the message box is remembered after a reload, and resets to the main agent if that agent is deleted or turned off.
  • •Dragging a file over the chat now highlights the message box itself, so it is clearer where the file will land.
  • •The project picker above the message box now reads "No Project" when you are working outside of a project, instead of "New Conversation".
  • •Improved performance when switching between conversations in the sidebar.
  • •Bold text in chat now renders at a slightly lighter weight so it sits better with the text around it.
  • •Fixed an issue where your saved settings and project list could be overwritten when the app failed to read its state file.
  • •Fixed an issue where permission settings accumulated duplicate entries, which made conversation files grow very large and slowed the app down.
  • •Fixed an issue where starting a new conversation could show a false crash recovery notice.
  • •Fixed an issue where a conversation started without a folder pulled in files and settings from unrelated projects you had open.
  • •A command that is killed when the app restarts is now reported as canceled instead of appearing to finish successfully.
  • •The agent now reports a clear error when it tries to read a binary file it cannot display, instead of returning unreadable content.
  • •A malformed tool call from the model now appears as a short "Invalid tool call" note rather than a failure.
  • •Fixed an issue where an error that arrived without any tool activity in the turn was dropped instead of shown in the conversation.
v1.2.6

Antigravity CLI: Session-scoped Remote Control, unlimited headless run timeouts, structured error reporting, and artifact text selection

18. September 2026

  • •Added Remote Control (start a connection via --remote-control startup flag or /remote-control slash command) to create a session-scoped remote connection for following and controlling your active terminal session from another device. Typing /remote-control off or closing the session automatically tears down the tunnel and unregisters the device from the active Remote Control session list.
  • •Changed the default timeout for headless (-p / --prompt) runs from 5 minutes to unlimited so long-running agent turns run until the response completes unless --print-timeout is passed explicitly, and enabled daemon background commands in headless GEMINI_API_KEY sessions so background servers stay running after the turn finishes.
  • •Improved headless (-p / --prompt) error reporting when a turn terminates on an agent or model API failure: the CLI now prints a structured AGY_ERROR: {...} JSON line on stderr with canonical status, HTTP or gRPC error code, retryability, and error ID (including HTTP status mapping for GEMINI_API_KEY SDK errors) and exits with code 3 instead of 1.
  • •Improved word-wise cursor movement (Alt+F, Alt+B, and Ctrl/Alt+Arrow keys) and word deletion (Ctrl+W, Alt+Backspace, and Alt+D) in the prompt editor to stop at punctuation boundaries instead of only whitespace, making it easy to step through or delete individual segments of file paths, URLs, and flags.
  • •Fixed turns triggered from a connected Remote Control session or running across secondary workspaces executing without the CLI session's active permission mode, cycle mode, and non-workspace file access grants.
  • •Fixed artifacts in the Remote Control companion rendering as plain .md file links instead of rich artifact cards when the CLI's application data directory differs from the default bundle name.
  • •Fixed text selection in the full-screen artifact viewer capturing the line-number gutter and trailing padding; dragging across the viewer now highlights and copies only the document text, with Ctrl+C to re-copy an active selection, Esc to clear it, and Shift+drag for native terminal selection.
  • •Fixed submitting /model while a turn is already running switching the active model immediately in the middle of the in-flight turn; the one-shot model switch now waits until the queued prompt begins executing so the running turn and any earlier queued prompts finish on the session's original model.
  • •Fixed file edits with diffs larger than 1 MiB exceeding the conversation storage limit and force-clearing the session; oversized edits now preserve their line-change statistics while omitting the raw diff body from stored conversation history and skipping the elided diff during /rewind reverts.
  • •Fixed /rewind conversation reverts and forks sweeping backward to an older workspace snapshot when recent snapshot commits were skipped, which could silently overwrite kept work; snapshot lookup now sweeps forward from the target step and falls back to step-by-step revert replay when no later snapshot exists.
v1.2.5

Antigravity CLI: Subagent roster prompt injection, background task names, credential invalidation handling, and thinking frame freeze fixes

17. September 2026

  • •Added automatic subagent guidance for custom agents defined in Markdown: agents that list invoke_subagent in their tools now receive the roster of available subagents and usage instructions in their system prompt, so they can actually delegate to the subagents they declared.
  • •Improved background task naming so a task keeps the name it was given when sent to the background; completion notifications and task lists now show that name instead of a generic auto-derived one.
  • •Fixed stale credentials lingering after the sign-in server definitively rejects them (for example a revoked or invalidated enterprise account); the CLI now signs you out and clears the stale tokens so you can sign in again cleanly, while temporary network failures, server errors, and rate limits never cost you your session.
  • •Fixed commands killed by an outside signal (for example when the whole session shuts down mid-run) being recorded as successful runs with exit code 0; such commands now finish as canceled while keeping the output collected before the interruption.
  • •Fixed the Interrupted hint only appearing when Esc was pressed locally in the terminal; it now appears whenever the current turn is actually cancelled, including when it is stopped from a connected remote session.
  • •Fixed the artifact viewer leaving markdown text stuck at its original width after a terminal resize; content now re-wraps to the new window size.
  • •Fixed in-progress thinking frames occasionally freezing permanently in the conversation history (for example showing Thinking... (31s) forever), including after a terminal resize or session reload.
v1.2.4

Antigravity CLI: Live skill reloading, interactive model search, mid-session thought signature fixes, and self-correcting tool validation

16. September 2026

  • •Added a /skills reload subcommand to asynchronously reload discovered skills and slash commands without restarting the session or blocking user input.
  • •Added an inline argument autocomplete dropdown when typing /model and a live fuzzy search bar inside the interactive /model picker to filter models by name or ID.
  • •Improved authentication error messages when OAuth tokens lack required scopes by including explicit instructions to run /logout and /login.
  • •Fixed mid-conversation model switching to Gemini thinking models failing with thought_signature validation errors when prior turns were generated by non-Gemini models.
  • •Fixed web search (search_web) failing with "no summary returned from GenerateContent" when the summarization model emits leading thought parts.
  • •Fixed agent turns terminating prematurely with NO_TOOL_CALL when a tool call fails schema validation, allowing the agent loop to return the validation error and self-correct.
  • •Fixed MCP tool schema augmentation mutating tool definitions in place, preventing strict argument validation failures.
  • •Fixed hooks.json configurations being silently dropped when customization token budget truncation is active.
  • •Fixed custom subagents losing configured exclude skill filters when runtime skill filtering is applied.
  • •Fixed subagent status and termination tracking under pubsub batch coalescing so killed subagents are not reset to active by late step updates.
  • •Fixed conversations with active turns or running daemon background tasks failing to auto-resume after a restart.
  • •Fixed terminal ERROR steps (such as failed tool executions or timeouts) being omitted from transcript.jsonl conversation logs.
  • •Fixed raw OSC 8 escape sequences rendering in Apple_Terminal over SSH connections by disabling unsupported hyperlink sequences.
v2.14.0

New Permissions System & Terminal and Git version control for Enterprise

15. September 2026

  • •Enterprise and Business accounts can now use the integrated terminal and Git version control in the sidebar.
  • •Conversations with very long histories now load faster and use less memory.
  • •Removed a redundant scan of your skills, rules, and other customization files that ran on every message, so messages are acknowledged faster.
  • •Settings, artifacts, and terminals opened in their own tab now load faster and no longer flash an empty container first.
  • •The agent's temporary scratch files are no longer watched, synced, or shown in the Artifacts panel and Review pane, so they no longer add noise to either.
  • •Reorganized the conversation menu so Archive and Delete sit together at the bottom behind their own divider, and added Share and Move to Group to the right-click menu.
  • •The command palette now opens over a dimmed backdrop with enter and exit animations, matching the other dialogs in the app.
  • •Renamed the global permissions section on the General settings screen to "Global Permissions" and the inherit option in project settings to "Inherit Global", so the names match the screen they point at.
  • •Pressing the new conversation shortcut on the home screen now toggles between your last-selected project and a standalone conversation instead of staying in standalone mode.
  • •Updated the Vesper dark theme with lilac and pink accents for keywords and operators, and dimmer comments.
  • •Fixed an issue where opening Settings, an artifact, or a terminal in its own tab could show a blank screen.
  • •Fixed a crash that could occur when the app failed to read your rule files.
  • •Fixed a memory leak where finished terminal commands held on to their full output for the rest of the session.
  • •Fixed an issue where the app could freeze for several seconds after reconnecting following a period of inactivity, which was most noticeable with a large number of conversations.
  • •Fixed an issue where subagents that had already finished could keep showing as running, which also left follow-up messages queued until restart.
v1.2.3

Antigravity CLI: Copy support for side questions, native backend tool-choice constraints, and subagent MCP inheritance

15. September 2026

  • •Added the /copy btw slash subcommand to copy the full text of the active /btw side-question response to the clipboard even when the response card is collapsed or scrolled, along with argument ghost hints when typing /copy.
  • •Improved tool-calling fidelity by translating per-request tool-choice constraints (any, required, and named functions) into native backend function-calling configurations.
  • •Fixed the /hooks command and hook inspection utilities omitting hooks bundled inside enabled plugins when listing active hooks.json configurations.
  • •Fixed custom subagents created with enable_mcp_tools: true receiving an empty MCP server list instead of inheriting the parent agent's configured MCP servers, and fixed declarative subagent configurations failing to resolve relative config file paths (relative_path_to_config) against the parent agent's directory.
  • •Fixed proactive feedback prompts triggering prematurely during the first few turns of a session or on transient, recovered tool errors.
v0.1.17

Antigravity IDE: Conversation compaction controls, forward-looking budget scopes, and tool output token truncation

14. September 2026

  • •Compaction Configuration: Adds CompactionConfig on AgentConfig to govern sliding-window conversation history compaction through an explicit token ceiling dial (token_threshold), deprecating CapabilitiesConfig.compaction_threshold.
  • •Forward-Looking Budget Scope: Adds BudgetScope.FORWARD_LOOKING to BudgetConfig to enforce model call and token budgets across newly resumed execution turns without counting previous historical usage.
  • •Tool Output Token Truncation: Exposes tool_output_truncation_config across agent configurations to cap token output volume initially for run_command executions in localharness.
  • •UsageMetadata Arithmetic: Supports standard Python arithmetic protocols on UsageMetadata, enabling scalar multiplications, scaling, and accumulating token counts using built-in sum().
  • •Tool Wrapper Metadata Preservation: Preserves function signature, module name, annotations, and original callable references via __wrapped__ when registering tools with ToolWithSchema.
  • •Interactive REPL Policy Flattening: Automatically flattens nested policy lists during interactive REPL upgrades so nested command authorization rules correctly upgrade to prompt the user.
  • •Terminal Command Sandboxing Guide: Added getting-started guide and references demonstrating OS-level command sandboxing using RunCommandConfig(enable_sandbox=True) paired with execution policies.
  • •Tool Call Deserialization: Fixed dropped tool arguments during tool call handling when incoming payloads provide structured dictionary arguments rather than serialized JSON strings.
  • •Local Step Trajectory Tracking: Fixed missing provenance metadata by forwarding trajectory_id from incoming tool calls to local connection execution steps.
  • •Dynamic Content Proto Resolution: Fixed an AttributeError when importing struct_converter in external environments missing internal protobuf definitions by resolving descriptor types dynamically at runtime.
v1.2.2

Antigravity CLI: Plugin MCP namespacing, conversation memory leak fixes, oversized file guardrails, and unsandboxed rule migration warnings

12. September 2026

  • •Improved the startup warning for deprecated unsandboxed permission rules across CLI, shared, and project configuration files to list each affected file path, up to five offending rules, and step-by-step instructions for migrating them to command rules.
  • •Improved /resume startup responsiveness when opening a large conversation history with a cold or stale summary cache.
  • •Fixed MCP servers bundled inside plugins colliding with each other or with user-configured servers in mcp_config.json when they shared the same server name by automatically namespacing plugin MCP servers as _.
  • •Fixed a memory leak where opening or scanning conversations left background step-cache eviction goroutines running for the rest of the session, significantly reducing memory usage after opening /resume or switching conversations.
  • •Fixed view_file attempting to parse non-UTF-8 binary files as text or loading files larger than 100 MB into the model context; unsupported binary formats and oversized files are now rejected with a clear error before overflowing the context window.
  • •Fixed Gemini API (GEMINI_API_KEY) sessions dropping model thinking blocks from prior turns on subsequent user messages and failing to propagate thought signatures on text and thought parts.
  • •Fixed artifact review failing to trigger when a directory name above .gemini/ matched a skipped path component such as scratch, prevented conversation forks and snapshot reverts from copying internal .system_generated/subagents and .system_generated/worktrees directories, and cleaned up subagent metadata records when deleting a conversation.
  • •Fixed deleted conversations being recreated as empty, schema-less SQLite database files when background queries reconnected after deletion.
  • •Fixed conversations launched without a workspace folder inheriting workspace paths and customizations from other open sessions.
v1.2.1

Antigravity CLI: Custom agent component exclusions, resilient model API retry handling, and conversation continuation fixes

11. September 2026

  • •Added support for excludeDefaultComponents: true in custom agent Markdown frontmatter, allowing custom agents to opt out of default prompt sections and built-in tools while preserving post-invocation hooks.
  • •Improved model API error resilience and diagnostics: transient genai.APIError failures (502, 503, 504, per-minute 429 rate limits, and mid-stream interruptions) automatically retry in-process with exponential backoff while preserving completed tool call outputs, and unrecovered 503 and 429 responses surface clear user-facing error messages.
  • •Improved MCP and provider tool schema validation to preserve open object schemas (such as {"type": "object"} or explicit additionalProperties: true) instead of rejecting undeclared arguments on schemas that allow them.
  • •Improved per-turn responsiveness and reduced memory usage when opening large conversations.
  • •Fixed --continue starting a brand-new conversation when launched from a subdirectory, after a crash, or while another session is open in the same workspace; it now falls back to the most recent non-empty conversation in the current workspace or its parent/child directories.
  • •Fixed full-screen blank flashes in inline mode when content first pushes to scrollback, at the end of a turn, or when clearing the screen with Ctrl+L.
  • •Fixed remote companion UIs connected to an interactive CLI session via Remote Control displaying an unauthenticated sign-in screen instead of the active session's signed-in state.
  • •Fixed the status line reporting the terminal sandbox as disabled when the session was launched with the --sandbox command-line flag.
  • •Fixed image zoom keys (Ctrl+= and Ctrl+-) and the footer zoom hint appearing in the artifact viewer when a Mermaid diagram is rendered in ASCII mode instead of as a Kitty image.
v1.2.0

Antigravity CLI: Hintergrunddienstregistrierung mit Remote-Control, erweiterter Halbseiten-Scrollbereich und Inhalts-Sicherheitsberichterstattung

10. September 2026

  • •Neu: Fernstart-, Fernstatus- und Fernstopp-Unterbefehle hinzugefügt, um CLI beim Systemdienst-Manager des Betriebssystems als Hintergrund-Daemon zu registrieren, der Abmeldungen und Neustarts überdauert. Enthalten: --name für benutzerdefinierte Instanz-Bezeichner und --session zum Limitieren auf die aktuelle Anmelde-Sitzung.
  • •Verbessert: Halbpaginansicht beim Scrollen über alle Ansichten (Alt-Modus, Diff-Ansicht und Detail-Panels, wobei Ctrl+D nicht mehr Exit-Prompts auslöst) und Hinzugefügt: Standard-Tastenkombinationen Shift+Pfeil Hoch und Shift+Pfeil Runter (navigation.half_page_up und navigation.half_page_down).
  • •Behoben: Probleme mit blocked prompts oder Modellantworten durch Content-Filter führten zu generischen Fehlermeldungen oder leeren Antworten; jetzt klare Stop-Gründe bei Verstößen anzeigen.
  • •Behoben: Temporäre Dateien in scratch/-Verzeichnissen haben rekursive Dateisystem-Watcher ausgelöst und Artefakt-Reviews sowie Checkpoints verunreinigt.
  • •Behoben: MCP-Server, die in globalen Plugins gebündelt sind, initialisieren beim CLI-Start nicht ordnungsgemäß oder aktualisieren laufende States nicht korrekt, wenn Plugins umgeschaltet werden.
  • •Behoben: Schema-Validierungsfehler bei Tools von Drittanbieter-MCP-Servern, deren Eingabeschemas zusätzliche Eigenschaften explizit ausschließen.
  • •Behoben: Redundante Dateisystem-Anpassungs-Discovery-Walks, die bei jeder Meldung liefen, obwohl keine Slash-Befehle aufgerufen wurden.
  • •Behoben: Spurenhaften ERRORs durch frühzeitiges Logging vor google.Init; betroffene Logs in cli.log und Ausgaben des Hintergrund-Daemons bereinigt.
  • •Fixed macOS background service lifecycle issues, legacy script cleanup, and paid Google Cloud project license availability for Remote Control.
  • •Fixed older conversations failing to load with unknown step type errors upon resumption.
v2.13.0

Introducing new Documents section and several UX enhancements for organization and control

9. September 2026

  • •External files such as Google Drive links, PDFs, and Office documents now appear in their own Documents section in the sidebar above Artifacts.
  • •Added a "Hide Whitespace Changes" option to the Review Changes overflow menu and file diff viewers to filter out whitespace-only edits.
  • •Code and data artifacts like SQL and JSONL files now open in a virtualized viewer with syntax highlighting and line numbers, keeping the app responsive when viewing large files and supporting inline comments.
  • •Commands executed in the sandbox now display a shield badge in the command header that links to sandbox documentation.
  • •Settings displays clear warnings with file paths and a copy button when configuration or project files fail to parse.
  • •Closing a side question now minimizes it to a button in the composer toolbar instead of clearing it. Clicking the button restores the panel so you can review previous side questions without retyping /btw, and a new delete button lets you remove questions you no longer need.
  • •Added a Cancel button and keyboard shortcuts (Ctrl+C on macOS, Ctrl+D on Linux and Windows) to interactive question prompts, letting you stop the agent and return to the chat input without submitting blank answers.
  • •Quote or add selected text to chat using Cmd+L or Cmd+I on macOS, or Ctrl+L and Ctrl+I on Windows and Linux. The text selection popup also adjusts button widths to fit shortcut labels.
  • •Agent scratch files now appear in a dedicated, collapsible Scratch Files section in the side panel rather than mixing into the primary Artifacts list. Turn cards also hide scratch files unless explicitly marked as user-facing.
  • •Unified sidebar section headers and spacing so row heights, collapse controls, and section gaps are consistent between pinned items and projects.
  • •Pin, unpin, and rename conversations directly from the command palette when a conversation is open.
  • •Added an option to archive the active conversation directly from the command palette, and updated archiving to navigate back to the home screen.
  • •Attached review comments now clear immediately from the input box and diff panes when you send a message, and restore if the message fails to send.
  • •Toast notifications stay on screen for 8 seconds instead of 5 seconds by default, giving you more time to read multi-line messages and error details.
  • •Adjusted spacing in the model quota panel and increased its maximum height to eliminate unnecessary scrollbars when viewing quota across multiple models.
v1.1.28

Antigravity CLI: Modell-API-Wiederholungsresilienz, headless Druckoptimierungen und detaillierte Tool-Genehmigungsabfragen

9. September 2026

  • •Verbessert: Resilienz gegenüber vorübergehenden API-Fehlern des Modells durch erneute Verarbeitung von Fehlern wie 503 Unavailable mit verlängertem exponentiellem Backoff, um kurze Aussetzer nicht zum Abbruch von Sessions führen zu lassen.
  • •Verbessert: Anmeldung und Startzeit durch Lesen der gültigen Identitätsdaten aus gespeicherten Anmeldedaten statt jedes Mal Netzwerkanfragen zu verwenden.
  • •Verbessert: Headless-Läufe (-p) beenden sich nun zügig, nachdem die endgültigen Antworten geliefert wurden, während auf Hintergrundaufgaben außerhalb des Daemons und Zeitpläne innerhalb von --print-timeout gewartet wird, und dabei laufende Hintergrunddaemons erhalten bleiben.
  • •Verbessertes Fehler-Reporting bei Headless-Läufen (-p): Fatalfehler werden nun auf stderr ausgegeben, mit stabilen Fehler-Markern und Hinweisen zur Trunkierung.
  • •Verbessert: Headless-Modus (-p) Latenz durch Beseitigung bis zu 200 ms Leerlauf-Verzögerung pro Runde und Weglassen ungenutzter Aufrufe des Modells zur Generierung von Gesprächstiteln.
  • •Verbessert: Tool-Approval-Anfragen zeigen nun konkrete Handlungen an (z. B. Run this command?, Allow access to this URL? oder Allow calling this tool?), Bearbeitung von Befehlen auf Befehle beschränkt. Falls ausgelöst durch Hooks oder projektübergreifende Dateien, erfolgt eine Reason-Erläuterung.
  • •Verbessert: Modell-Auswahl-Überprüfung durch Protokollierung von Alias-Auflösungen, Mapping von --effort-Varianten und veralteten Modell-Ersetzungen in cli.log.
  • •Neu: Verhalten des --print-timeout im Headless-Modus angepasst: Teil-Ausgabe wird zurückgegeben und mit Warnung auf stderr erfolgreich beendet statt Timeout-Fehler auszulösen.
  • •Changed default URL fetching permissions to prompt for approval by default before reading external URLs unless access has been pre-granted.
  • •Fixed plugin reinstalls leaving behind files deleted from the source by replacing managed plugin directories cleanly, preventing self-directory installation corruption, and cleaning up disabled state entries from config.json upon uninstall.
  • •Fixed plugin-defined MCP servers resolving relative working directories incorrectly by scoping them directly to the plugin's root directory.
  • •Fixed a startup race condition where subagents could fail to discover tools from MCP servers that were still initializing.
  • •Fixed subagents occasionally appearing stuck in running states in large conversations, resolving delays delivering queued idle messages.
  • •Fixed a memory leak where finished terminal command executions retained internal state throughout long-running sessions.
  • •Fixed headless (-p) runs hanging indefinitely on interactive implementation-plan approval prompts by proceeding through plan review automatically in non-interactive mode.
v1.1.27

Antigravity CLI: Mid-conversation model execution, benutzerdefinierte Agent-Subagent-Abhängigkeiten und MCP-Parameter-Validierung behoben

5. September 2026

  • •Neu: /model, das eine einzige Aufforderung auf einem anderen Modell ausführt und dann die Sitzung wieder dem ursprünglichen Modell zurückgibt, sodass während einer Unterhaltung ein anderes Modell konsultiert werden kann, ohne das gespeicherte Standardmodell zu stören.
  • •Neu: Ein Feld conversation_title im JSON-Payload für benutzerdefinierte Statuszeile und Fenstertitel-Skripte hinzugefügt, damit der Name der aktiven Unterhaltung angezeigt wird und Umbenennungen sofort übernommen werden.
  • •Neu: Eine Agentenliste im benutzerdefinierten Agenten-Frontmatter hinzufügen, damit ein Agent die abhängigen Subagenten deklarieren kann. nach denselben workspace-relativen, absoluten und agenten-relativen Pfadregeln wie bei Skills.
  • •Verbessert: /model zeigt beim Tippen einen [model]-Argument-Hinweis, sodass die Argumente der Befehle ohne Hilfe geöffnet werden können.
  • •Verbessert: Das /settings-Panel erklärt jede Verbalität (Verbosity) beim Hervorheben, beschreibt, ob Tool-Aufrufe, Befehle und Gedankengänge vollständig angezeigt oder zu einer Token-Metrik-Tabelle zusammengefasst werden.
  • •Verbessert: Anmelde- und Google-Cloud-Einrichtungsbildschirme verwenden jetzt die standardisierten CLI-Schlüsselhinweise, richten sich nach Farbschema und Tastaturbelegung und entfernen den Navigationshinweis, wenn nur eine Option verfügbar ist.
  • •Behoben: Das Modell-Reporting gibt jetzt an, dass das Modell bereits gesetzt ist, wenn in der Sitzung bereits dieses Modell verwendet wird, statt von einem Wechsel zu berichten, der nicht stattgefunden hat.
  • •Behoben: Auf Tool-Aufrufe zu MCP-Servern, die Argumente akzeptierten, die dem eigenen Schema des Servers nie deklariert waren; ein erfundener Parameter wird jetzt abgelehnt und korrigiert statt still zu fallengelassen.
  • •Fixed headless runs with -p silently skipping tool actions they were not permitted to take, which now end with a notice naming the refused actions and report them as denied_actions in the JSON output.
  • •Fixed ? in Vim Normal mode opening the shortcuts panel while the prompt already contained text, so character-search and replace operators such as f?, t?, and df? work; ? still opens shortcuts on an empty prompt.
  • •Fixed /model with no argument doing nothing when run before the session had finished starting up, so it now opens the model picker.
  • •Fixed print mode (-p) exiting before the session finished shutting down, which could drop the run's trailing conversation history before it reached disk.
  • •Fixed a redundant approval prompt when the agent reads or writes another conversation's artifact files after you have already allowed access to files outside your workspace.
v1.1.26

Antigravity CLI: Halbfester Seiten-Scrollbereich der Artefakt-Ansicht, Gruppierungseinstellung für Workspace-Auswahl, und Flush von SQLite WAL

4. September 2026

  • •Neu: Halbseiten-Scrollen im Artefakt-Viewer mit Ctrl+D und Ctrl+U hinzufügen
  • •Neu: pickerGrouping-Einstellung zu /config und settings.json hinzugefügt, um Standardansicht der Gesprächsliste in /resume festzulegen (Flachansicht oder gruppiert nach Workspace).
  • •Verbessert: Terminal-ASCII-Diagrammdarstellung für Mermaid-Flowcharts im Artefakt-Viewer und in der Konversation
  • •Verbessert: Logout-Ausführung durch Direktspeicherung der Token-Entfernung in die Dateispeicherung, wenn Keyring-Speicherung umgangen oder nicht erreichbar ist.
  • •Neu: Unausgewählte Modellfamilien im interaktiven /model-Picker standardmäßig auf mittleren Abstraktionsgrad/Begründungsaufwand eingestellt statt auf geringer.
  • •Behoben: Subagent-Aufgaben fordern unerwartet Tool-Berechtigungen im Always-Proceed-Modus während der Ansicht des Subagenten-Details-Fensters.
  • •Behoben: verwaiste Git-Worktrees, die sich durch das Beenden von Subagenten oder das Löschen von Unterhaltungen auf der Festplatte unter.system_generated/worktrees ansammeln.
  • •Behoben: SQLite-Datenbank-WAL-Checkpointing beim CLI-Ausstieg, sodass abschließende Metadaten der Sitzung vor dem Herunterfahren auf der Festplatte geschrieben werden.
  • •Fixed customization discovery logging spurious errors when accessing temporary directories or paths outside the active workspace.
v2.12.2

Gemini 3.8: Flashen mit ADC in AGY Enterprise

3. September 2026

  • •Neu: Enterprise-Nutzer können ab sofort Gemini 3.8 Flash über ADC für agentische Aufgaben auswählen und nutzen.
v1.1.25

Antigravity CLI: Arbeitsbereich gruppierte Resumen-Ansicht, Gemini 3.8 Flash-Unterstützung und Markdown-Agenten-Ambient-Anpassung-Vererbung

3. September 2026

  • •Neu: Opt-in workspace-gruppenbasierte Ansicht im /resume-Konversationsauswahlfenster, mit der Benutzer zwischen flacher Liste und nach Verzeichnissen gruppierten Unterhaltungen mit Ctrl+F wechseln können.
  • •Neu: Gemini 3.8 Flash in das Modellkatalog aufnehmen, wenn eine GEMINI_API_KEY-Verbindung besteht.
  • •Neu: Angepasste benutzerdefinierte Agents übernehmen standardmäßig kontextabhängige Fähigkeiten, Regeln und Subagents und stimmen sich damit an die Konfiguration der Standard-Agents an.
  • •Behoben: MCP OAuth-Authentifizierung schlägt fehl, wenn Autorisierungsserver längere Authorization Codes als 1024 Zeichen zurückgeben.
  • •Behoben: Abgleich und Gruppierung von Skill-Pfaden im /skills-Panel unter Windows, bei denen Host-Pfadtrenner (\) globale und Arbeitsbereich-Skills falsch klassifiziert wurden.
  • •Behoben: Auflösung relativer Pfade in Markdown-definierten benutzerdefinierten Agents. relative Sub-Agent- und Plugin-Pfade werden nun korrekt im Verzeichnis der Agenten-Definition aufgelöst.
  • •Behoben: Duplizierte Berechtigungen kumulieren sich in den Konfigurationseinstellungen über Sitzungsneustarts hinweg und Unteragentenaufrufe hinweg.
  • •Verbessert: Stabilität durch optimierte Worker-Tool-Konfigurationen; Fehlschläge beim Boosting zur Laufzeit behoben.
  • •Fixed a fatal nil pointer crash in the agent runtime caused by background summary updates attempting to read cached steps after trajectory closure.
  • •Hardened Remote Control reverse-tunnel routing.
v2.12.0

retry

2. September 2026

  • •Neu: Abschnitte von Antigravity-Antworten zum Zitieren als Kontext für Folgeaufforderungen hervorheben können.
  • •Neu: Den /boost-Befehl eingeführt, um die Denkleistung durch eine mehrstufige reasoning-Pipeline zu erhöhen.
  • •Neu: Allgemeine Einstellungen zeigen jetzt, welche deiner Projekte eine Einstellung überschreiben, mit direktem Link zu den jeweiligen Projekteinstellungen. Öffnet man die Einstellungen, wird die aktive Projekts-Einstellung angezeigt, sofern Überschreibungen vorliegen.
  • •Anzeige- bzw. Terminal-Layout bleibt beim Neuladen des Fensters erhalten.
  • •Neu: Die Seitenleiste mit dem Frage-Antwort-Panel erhält Zurück- und Vorwärtssteuerelemente zum Durchblättern der gestellten Fragen in einer Sitzung.
  • •Neu: Nachricht jetzt senden, während Diktat läuft. kein Stopp der Aufnahme mehr nötig.
  • •Neu: Auswählen eines Bildausschnitts per Ziehen zum Kommentieren aktualisiert die Auswahl sofort beim Loslassen der Maustaste.
  • •Neu: Drag-and-Drop verbessert und Unterstützung zum Anhängen von Audiodateien an Unterhaltungen erweitert; Problem behoben, bei dem WebM-Dateien fälschlicherweise als Audio behandelt wurden.
  • •Fixed an issue that could cause authentication errors part way through a turn on slower network connections.
  • •Permission prompts again describe the specific action being requested instead of showing a generic title.
  • •Fixed an issue where subagents running side by side disappeared from the conversation when one of them reported progress.
  • •Fixed an issue where an extension or automation could briefly show an unavailable error after being restarted.
  • •A failing custom hook no longer ends the session; it is now reported as an error and the conversation continues.
  • •Fixed the connected tool server configuration dialog overflowing its window.
  • •Fixed the built-in workflow migration helper reporting wrong paths and moving the wrong files on Windows.
v1.1.24

Antigravity CLI: MCP-Panelnavigation verbessert, mehrzeilige Kommentare in mcp_config.json unterstützt und Headless-Stream-Aufräumarbeiten

2. September 2026

  • •Verbessert: MCP-Panel-Navigation. Auf- und Abwärtstasten navigieren strikt zwischen MCP-Servern und Plugins; Links- und Rechts-Tasten durchlaufen die verfügbaren Aktionen; Escape beendet den Aktionsmodus bzw. das Panel.
  • •Behoben: doppelte Agenten-Einträge im /agents Picker, wenn derselbe Agent aus mehreren Konfigurationsquellen oder Plugins entdeckt wird.
  • •Behoben: Tool-Initialisierung und Start des Agents scheiterten, wenn aus einem nicht zugänglichen oder gelöschten Arbeitsverzeichnis gestartet wurde. Abhilfe durch Verwendung absoluter URI-Schemata für In-Memory-Parameter-Schemata.
  • •Behoben: Headless CLI-Aufrufe mit gepiepter Standardausgabe oder Standardfehler führen beim Beenden nicht mehr zu Hänger, indem FD_CLOEXEC auf den beibehaltenen Streams gesetzt wird. Kindliche Prozesse schließen dadurch die Pipes des Aufrufers nicht off.
  • •Behoben: MCP-Konfigurations-Parsing schlägt fehl, wenn mcp_config.json einzeilige (//) Kommentare, mehrzeilige (/ /) Kommentare oder abschließende Kommas enthält.
  • •Behoben: Verwaiste Annotationsdateien stapeln sich nicht mehr im lokalen Speicher beim Löschen von Gesprächen.
  • •Behoben: Diskussionen mit aktivem /goal erzwingen und ungewollte Tool-Aufrufe auslösen.
v1.1.23

Antigravity CLI: Modell-Autovervollständigung verbessert, Subagent-Streaming optimiert und MCP-Toolauflösung behoben

1. September 2026

  • •Verbessert: /model-Autovervollständigung akzeptiert den vorgeschlagenen Modellnamen als Ghost-Text per Tab.
  • •Verbessert: Subagent-Streaming-Overhead durch einmalige Übermittlung der Subagenten-Trajektorien-Metadaten pro Subtrajekt statt bei jedem Schritt.
  • •Behoben: Befehle mit Unterbefehlen (z. B. Modelle oder Agents) hängen beim vererbten Standardeingabestrom.
  • •Behoben: CLI-Crashed durch Prompt-Hooks verhindert, indem Hook-Paniken abgefangen und nil-Completion-Konfigurationen in Modell-Anfragen abgelehnt werden.
  • •Behoben: Aufrufe von Tools und deren Ergebnisse beim Rekonstruieren des Anfrage-Verlaufs für Gemini-Modelle IDs von tool-call-Blöcken ausgelassen.
  • •Behoben: Tool-Berechtigungsabfragen für native- und MCP-Toolaufrufe zeigten generische Titel statt menschenlesbarer Aktionsbeschreibungen.
  • •Behoben: Nicht konfigurierte oder unterbrochene Google-Cloud-Authentifizierung führte zu defekten Chats statt der erneuten Aufforderung zur Anmeldung.
  • •Behoben: Vorübergehende Authentifizierungsfehler durch Uhr-Abweichungen wurden behoben, indem Browser- und WIF-OAuth-Tokens proaktiv 5 Minuten vor Ablauf aktualisiert wurden.
  • •Fixed instructional cycle mode placeholders reappearing in the input prompt after typing and clearing content.
  • •Fixed subagents defined with enable_mcp_tools=true failing with unknown tool errors by ensuring the MCP dispatcher is available to the subagent executor.
  • •Fixed prompts submitted immediately after login being discarded by queueing them until background account verification finishes.
  • •Fixed cancelled or killed subagents remaining stuck in the "Running" state in the UI task drawer across ancestor conversations.
  • •Fixed parsing MCP JSON configuration files on Windows when saved with a UTF-8 byte order mark (BOM).
v0.1.16

Antigravity IDE: Standard Gemini 3.8 Flash-Modell-Upgrade, leichter Agenten-Konfiguration und Vertex AI Express-Modus

31. August 2026

  • •Standardmodell-Update auf Gemini 3.8 Flash: Das Standardmodell für neue Agents und Konfigurationen wurde auf gemini-3.8-flash aktualisiert, mit höherer Begründungsqualität und stärkerer Aufgabenleistung. Entwickler können diesen Standard durch Angabe des Modells in der Konfiguration überschreiben.
  • •Optimierte Konfiguration des Lightweight-Agent: Neue Methode.lightweight() in Agent-Konfigurationen (einschließlich LocalAgentConfig und LiteRTAgentConfig) wendet voreingestellte Optimierungen für kleinere, lokal laufende Modelle an, indem minimale Tool-Sets, minimaler Prompting-Aufwand und Deaktivieren von Hintergrund-Subagenten konfiguriert werden.
  • •Neu: Vertex AI Express Mode (API Key-Unterstützung) ermöglichen die Verbindung zu Vertex AI via Express-Modus durch Angabe eines API-Keys direkt in LocalAgentConfig(vertex=True, api_key=„...“), erleichtert die Authentifizierung ohne vollständige GCP-Projekt-/Standort-ADC-Einrichtung.
  • •Unterstützung für Callable Class- und Dataclass-Tools: Der ToolRunner des SDK inspiziert und führt Werkzeuge aus, die als aufzurufende Klasseninstanzen (Funktoren) bzw. Dataclass-Methoden definiert sind, wodurch benutzerdefinierte Tool-Implementierungen flexibler genutzt werden können.
  • •Neu: OS Sandbox Opt-in für Befehle Hinzugefügt: addon enable_sandbox-Feld zu RunCommandConfig, um Entwicklern das Ausführen von Terminalbefehlen in einer OS-Ebene-Sandbox-Umgebung zu ermöglichen
  • •Neu: ToolWithSchema und öffentliche Callables unterstützen nun sowohl positionsbasierte als auch Schlüsselwort-Argumente (*kwargs), gemäß üblicher Python-Funktionsaufrufe.
  • •Unterstützung für genai.Content-Medien: Einführung eines StructConverters zur Verarbeitung von Medienblöcken aus genai.Content-Objekten im SDK, enabling reichhaltigere multimodale Interaktionen.
  • •Neu: MCP-Abhängigkeitskompatibilität erweitert: Das SDK unterstützt jetzt sowohl mcp>=1.0 als auch mcp<3.0 Abhängigkeiten.
  • •Stop Hook Integration: Agents now support the StopHook lifecycle hook, triggered when an agent's execution is externally stopped, enabling custom cleanup and resource logging actions:
  • •Interactive Example Agent Behavior: Fixed interactive SDK examples (interactive_cli.py, human_in_the_loop.py) to consistently configure AgentBehavior.INTERACTIVE.
  • •Tool Runner stdio MCP Environment: Resolved an issue in exported SDK examples where stdio MCP servers failed by ensuring child processes use sys.executable within the active virtual environment.
  • •Step Token Usage Deprecation: Deprecated Step.usage_metadata in favor of consolidated turn-level (ChatResponse.usage_metadata) and session-level (agent.conversation.total_usage) reporting.
v1.1.22

Model-Auswahlbefehle: Argumente, reasoning effort Auto-Vervollständigung, begrenztes Retry-Handling und UI-Performance verbessert

27. August 2026

  • •Neu: Ein /model-Argument hinzufügen, um Modelle per Name, Slug oder Label zu wechseln und Standardwerte in einem Schritt festzulegen. inklusive Inline-Ghost-Text-Vervollständigung.
  • •Verbessert: Der /effort-Hinweis ergänzt dynamisch den eingegebenen Text statt eines statischen Platzhalters anzuzeigen.
  • •Verbessert: Artefaktverarbeitung in Sitzungen, die viele Dateien erzeugen, indem Burst-Ereignisse des Dateisystems zu einem einzigen Neustart-/Rescan-Vorgang zusammengeführt werden.
  • •Behoben: Selektiven Reasoning-Aufwand-Konfigurationsparameter für Gemini 3.1 Pro und Gemini 3.5 Flash beim Authentifizieren über Gemini API-Schlüssel.
  • •Behoben: Kontinuierliche Neudarstellung der Benutzeroberfläche, wenn das Tasks-Panel oder das Subagent-Detail-Panel geöffnet war, obwohl keine aktiven Tasks vorhanden waren; CPU-Idle-Verbrauch deutlich reduziert.
  • •Behoben: Laufende Subagenten-Elapsed-Timer frieren nicht mehr ein, während der übergeordnete Agent im Wartezustand ist.
  • •Behoben: temporäre HTTP-502-Fehler Bad Gateway, die Läufe abbrachen, durch automatische Wiederholungslogik mit Backoff.
  • •Beheben: Selbstständige Subagenten, gestartet aus unausgestatteten Sitzungen, lösen Konfigurationen erneut von Grund auf neu auf und weichen dabei von den elterlichen autonomen Ausführungsmodi ab.
  • •Fixed Windows file deletion sharing violations by retrying deletions with backoff before failing.
  • •Fixed headless daemon startup banner printing unsupported localhost browser URLs to service manager logs.
  • •Fixed the built-in migrate-workflows skill to properly handle Windows path separators and home directories.
v2.11.0

Generatives UI und UI-Verbesserungen

26. August 2026

  • •Neu: Unterstützung für das direkte Referenzieren und Einbetten externer Dateien mittels @path/to/file in AGENTS.md und benutzerdefinierten Regel-Dateien hinterlegt.
  • •Neu: Unterstützung für das Rendern von YAML-Frontmatter als formatierte Metadata-Karte in Markdown-Dateien und Artefakten
  • •Neu: Unterstützung für KaTeX-Formeln, Chart.js und Plotly-Charts in generativen UI-Widgets und Artefakt-Vorschauen
  • •Neu: Terminale nebeneinander im Terminal-Panel dank Split-Button oder Tastenkürzel (Ctrl+Shift+5 auf Linux/Windows, Cmd+\ auf Mac) trennbar; verschiebbare Trennergröße sowie Baum-Indikatoren in der Terminalliste.
  • •Neu: Darcula-Theme-Vorgabe hinzugefügt und Farbschema für Syntax-Hervorhebung ergänzt
  • •Neu: Unterstützung zum Anzeigen spezifischer Seitenbereiche (StartPage, EndPage) in mehrseitigen Dokumenten (z. B. PDFs) sowie Festlegung der Bildauflösung (MediaResolution in view_file).
  • •Neu: Unterstützung für Hover-Vorschau-Karte beim Skills-Chip im Chat
  • •Neu: Unterstützung zum Definieren eines rules: key im Frontmatter des benutzerdefinierten Markdown-Agenten, um spezifische Regeldateien direkt an einen Agenten zu binden.
  • •Added support for discovering custom skills, agents, and rules defined in skills.json, agents.json, and rules.json configuration files located within project subdirectories.
  • •Several overall UX improvements for performance, accessibility, and project configurations.
  • •Fixed an issue where slash commands unavailable to a selected custom agent were incorrectly shown in the command menu.
  • •Fixed an issue where image attachments in the message composer were lost when switching between conversations.
  • •Fixed an issue where Ogg audio and video files were assigned generic MIME types that could cause model errors.
  • •Fixed an issue where custom color themes with transparency or alpha channels caused UI text to render in red.
  • •Fixed an icon mismatch where the title bar overflow menu displayed a generic terminal icon instead of matching the tab icon in the side panel.
v1.1.21

Antigravity CLI: Sprachdictation, eingebautes Ripgrep-Code-Suchen, Statusleisten-Kostenmetriken und UTF-8-Encoding-Fehlerbehebungen

26. August 2026

  • •Hinzugefügt: Sprachdictation, um Sprache direkt in den Prompt zu transkribieren (F5- und /voice-Toggle). Neu: mic-serve-Unterbefehl zum Weiterleiten eines lokalen Mikrofoninputs über SSH. Verbessert: Quotenmeldungen bei gleichzeitigen Streams.
  • •Neu: Kosteneinheit in der Statusleisten-Datenmodell eingeführt; schätzt laufende Token-Ausgaben und gibt ungerundete Sitzungs-Kosten an.
  • •Verbessert: Agenten-Code-Suche durch Einbindung eines eingebetteten ripgrep-Binärdatei statt externer System-Binärdateien, steigert Suchleistung und plattformübergreifende Konsistenz.
  • •Verbessert: Der Always-Proceed-Berechtigungsmodus genehmigt MCP-Tool-Aufrufe und Seitenlesungen automatisch, ohne Abfrage.
  • •Verbessert: Vorgeschlagene Always-Allow-Berechtigungen für Script-Runner (npm run, yarn, pnpm, cargo run) durch Einschränkung der Freigaben auf konkrete Script-Namen.
  • •Verbessert: Diskussionstitel automatisch bei Erstellung erstellen, aussagekräftige Sitzungsnamen im Resumé-Auswahlmenü liefern.
  • •Verbessert: Fehlersignale bei MCP-Servern mit Google-Credentials; konnte Application Default Credentials (ADC) nicht finden.
  • •Behoben: Beschädigte Bearbeitungen von Dateien in Dokumenten mit nicht-ASCII-Zeichen (CJK, Akzentbuchstaben, Emoji) durch ungenaue Offset-Übereinstimmungen, die ungültiges UTF-8 erzeugten.
  • •Fixed sessions stalling mid-response when a tool result or file diff contained invalid UTF-8.
  • •Fixed file writes being reported as failures after successfully saving to disk, preventing unnecessary agent retry loops on large files and notebooks.
  • •Fixed precedence resolution so explicitly configured skill and plugin paths take priority over automatically discovered customizations during name collisions.
Zeige 30 von 136