Why build complex runtime infrastructure when you can spin up secure, sandboxed Linux environments to execute agentic workflows with a single API call?

Google has launched the Managed Agents API within the Gemini API and Gemini Enterprise Agent Platform. Managed Agents API abstracts away agent execution, allowing you to deploy autonomous agents instantly. With a single API call, developers can deploy production-ready agents that execute code, browse the web, and manage files in isolated, secure Linux sandboxes. By shifting the agent harness, state persistence, and environment compute directly into Google infrastructure, developers can build complex autonomous agentic workflows without managing underlying infrastructure.

• Production-Grade Harness: Built on Google internal agent architecture co-optimized with Gemini models.

• Zero Infrastructure Management: Provisions secure OS-level Linux sandboxes, tools, and execution loops instantly.

• Stateful and Persistent: Configure environment dependencies, data sources, and repos once, then invoke by ID.

• Enterprise Governance: Seamlessly integrates with Google Cloud enterprise controls and security policy frameworks.

To learn more:

Google Blog: Managed Agents in the Gemini API

Gemini Enterprise Agent Platform Documentation

• After a meeting, we’ll automatically upload these files to a new folder in the host’s My Drive called “Google Meet.”
• Within that “Google Meet” folder, these files will be automatically organized into subfolders for each meeting. Files from different instances of a recurring meeting will share one folder.
• Any meeting attendees with access to the meeting files will see shortcuts to these source files in their “Google Meet” Drive folders too..

Previously, these files were only uploaded to the host’s My Drive, not the attendees’, and the files weren’t organized by meeting.

Note: Shortly after we roll out this change, we’ll automatically move the existing “Meet Recordings” folder into the new “Google Meet” folder and rename it “Legacy Meet Recordings.” Users may see both “Meet Recordings” and “Google Meet” in their Google Drive for a brief period during this transition.

For more information, see Create and edit documents and slides in Canvas.

A Gemini Enterprise app administrator must turn on the Enable canvas toggle in the web app feature management settings to let users use it. For more information about feature controls, see Manage features on the web app.

To request access to Gemini 3.6 Flash in the US multi-region, contact your Google account team.

Claude Opus 5 on Google Cloud is Anthropic's most advanced Opus model, powering long-running agents while delivering improvements in coding and professional work.

*Additional reading

🚀 Google Antigravity (Suite, CLI, & SDK)

The foundation for building and running automated development workflows has received significant feature updates to streamline agent control and sandbox execution.

• Antigravity Desktop Queued Messages: The desktop app now supports queued messages, offering developers more controlled message flow execution, alongside performance, stability, and usability improvements.

• CLI Plan Modes: You can now cycle through agent execution modes (such as plan or accept-edits) and design modifications ahead of execution with the new /plan mode. The Antigravity CLI also received several critical stability fixes.

• ADK Python Framework Upgrade: The Python framework adds native support for running in Cloud Run sandboxes and drastically expands Agent Registry capabilities.

• Knowledge Catalog Connector Expansion: The catalog adds support for automated metadata imports from SQL Server and PostgreSQL sources to keep technical assets indexed.

☁️ Cloud Run, GKE, and Infrastructure

We have introduced major GKE scaling improvements and streamlined registry import protocols to make running containers in production safer and more customizable.

Our fastest, most cost-effective 3.5-class model, delivering 350 output tokens per second.

*Additional reading

Our workhorse model that delivers better coding, knowledge work, and multimodal performance.

*Additional reading

Agents CLI has officially reached version 1.0 and is now Generally Available (GA). It is an Agent Development Lifecycle toolchain that equips coding assistants, such as Antigravity CLI, Claude Code, and Codex, with the skills and commands needed to seamlessly build, evaluate, govern, and deploy enterprise-grade AI agents on Google Cloud.

Be an agent building, scaling, governing and optimization expert without having to read the docs.

Why it matters:  

The GA release is designed to help developers accelerate from local prototyping to production deployment. Based on community feedback since Google Cloud NEXT, the 1.0 release introduces several critical capabilities:

• Rock-Solid Deployments: Smarter redeploys on Agent Runtime and Cloud Run that preserve your existing deployment specs instead of resetting them.

• New RAG Recipes: A shift toward flexible, clone-and-study RAG samples located in google/adk-samples, allowing you to easily adapt and build exactly what you need.

• Enterprise-Ready: Native Gemini Enterprise registrations and secure Cloud Trace spans that keep sensitive LLM prompts out of your observability traces.

• Simplified DX: Consolidated .env configs, robust evaluation commands, and refreshed skills to give your coding agents the best possible context for generating and evaluating code.

*Additional reading

The code-optimization agent, built on Gemini, is now generally available to every Google Cloud customer on the Gemini Enterprise Agent Platform.

# What you feed it

Two inputs. A seed program with the code segments you want rewritten marked out, and an evaluator: a deterministic script that runs on your side to compile, test, and score each candidate. AlphaEvolve then loops through generating variants, scoring them, and evolving the winners. Google frames the workflow as define, measure, optimize, apply, and aims it at logistics, semiconductors, genomics, HPC, and financial services. IDE integration covers Antigravity and Claude Code.

• What it does: Fill with Gemini allows users to easily generate text, summarize information, categorize data, and analyze sentiment at scale directly within spreadsheet cells, eliminating the need for complex formulas.

• 
  

• New Languages: Mandarin, Dutch, Malay, Hebrew, Polish, Turkish, Czech, Indonesian, Swedish, Danish, and Norwegian. (This is in addition to the previously supported English, Spanish, Portuguese, Japanese, Korean, French, Italian, and German).

• 
  

• Rollout: Gradual rollout (up to 15 days for visibility) starting on July 7, 2026 for both Rapid and Scheduled Release domains.

• 
  

• Requirements: End users must have Workspace smart features enabled to use it. It can be triggered by dragging a selection or selecting empty cells.

• 
  

• Availability: It's available for users with specific Business, Enterprise, Consumer, and Education plans, as well as those with AI Add-ons. (Note: Starting July 15, 2026, users with AI Expanded Access licenses will get higher usage limits).

Google launched the Interactions API as its primary interface for Gemini models and agents, and expanded Managed Agents in the Gemini API. If you build on Gemini, plan a migration path from bare `generateContent` calls toward this agent-centric surface; Google is signaling where the API investment goes from here.

*Additional reading

DeepMind released Gemini 3.5 in May under the banner "frontier intelligence with action." Flash is the variant you can use today. Google positions it for agents and coding, and in June added computer use, so the model can drive a desktop session on its own. Published numbers: 76.2% on Terminal-bench 2.1 and 55.1% on SWE-Bench Pro. Gemini 3.5 Pro is "coming soon."

*Additional reading

Welcome to this month's Gemini Enterprise newsletter. Over the past several weeks, Google Cloud has rolled out major updates focused on observability, collaborative security, and expanding our connected data ecosystem

Whether your goal is to securely monitor thousands of AI agents or integrate your existing ServiceNow workflows, these new features ensure your organization can confidently scale AI safely

Below is a breakdown of the latest updates, their release status, and how to get started.

🌎 Global Reach & Compliance

📍 Expanded Regional Support (Japan & United Kingdom)

• Release Status: GA (with Allowlist)

• Details: In-region support launched for Japan (asia-northeast1) and the United Kingdom (europe-west2), enabling at-rest Data Residency, Machine Learning Processing, and native Gemini 3.5 Flash. Contact your account team for access.

🛠️ Agentic Customization & Workflows

👥 Share Custom Agents with Google Identity Groups

• Release Status: General Availability (GA)

• What it is: End users who build custom AI agents via the Agent Designer can now share them directly with Google Identity groups.

• How to get started: Administrators must first enable this sharing feature inside the Gemini Enterprise app feature controls.

🗂️ Import Agents and MCP Servers via Agent Registry

• Release Status: General Availability (GA)

• What it is: You can now access Agent Registry directly from within Gemini Enterprise to select A2A (Agent-to-Agent) assistants and custom Model Context Protocol (MCP) servers.

🔗 Connected Workspace Ecosystem

🛠️ ServiceNow Data Store Action Integration

• Release Status: General Availability (GA)

• Details: Securely connect ServiceNow as a data store to search incidents, tasks, and knowledge base articles, and execute actions directly within Gemini Enterprise.

📈 Platform Core & Observability

🔍 Deep Observability for Gemini Enterprise

• Release Status: General Availability (GA)

• What it is: A comprehensive telemetry suite that gives administrators and developers deep visibility into agent performance and system usage.

🎨 Gemini 3.1 Flash Image & Gemini 3 Pro Image Support

• Release Status: General Availability (GA)

• What it is: Image models are now natively supported in the Gemini Enterprise app.

🚀 Get Involved & Next Steps

• Ready to roll out these updates? Head to your Google Cloud Console to configure your new ServiceNow connections and review your Agent Gateway egress policies.

• Learn More: Check out the official Google Cloud documentation for a deep dive into configuring observability settings.

🤖 GKE & AI Orchestration GKE Agent Sandbox: Google introduced a new GKE add-on designed for the agentic AI era. It manages isolated, stateful, and single-replica workloads optimized for AI agent runtimes, allowing for the deployment of 300 secure sandboxes per second per cluster. 🔗 Accelerating the agentic era with GKE

GKE Pod Snapshots: Designed for AI inference and long-initialization workloads, Pod Snapshots allow GKE to save a pod's memory and file system state. New replicas can now restore from these snapshots, drastically reducing startup latency. 🔗 GKE Pod snapshot

GKE Capacity Buffers Preview: GKE introduced Capacity Buffers to combat scale-up latency. By proactively declaring tiers of active or standby (suspended) nodes, critical applications and AI workloads can scale instantly without waiting for new VM provisioning. 🔗 About capacity buffers

GKE Hypercluster Private GA: Redefining the scalability ceiling, GKE Hypercluster allows a single Kubernetes-conformant control plane to manage up to a million chips distributed across 256,000 nodes and multiple regions, creating a unified capacity reserve. 🔗 Scaling foundational models with GKE hypercluster

⚙️ Compute & Networking New Compute Series Previews: Google announced the upcoming Q3 2026 previews for Z4M VMs and new bare metal instances, designed to provide the highly fluid compute needed for unpredictable bursts in agentic planning and reinforcement learning tasks. 🔗 What's new in compute at Next '26

Cross-Cloud Network for Agentic Enterprises: Google highlighted innovations in its Cross-Cloud Network to support massive-scale inference and reinforcement learning. The network is positioned as the critical integration layer to connect and protect agents, AI services, and core applications. 🔗 What's new in cloud networking at Next '26

💾 Storage & Infrastructure Smart Storage Layer: Google Cloud Storage introduced the new "Rapid" family for high-performance object storage and a dynamic tier for Managed Lustre. The storage layer now integrates Google's AI models to provide automated metadata annotation. 🔗 Next '26 storage announcements

Intent-Based Autoscaling in GKE: Moving beyond basic CPU/memory metrics, GKE now supports native custom metrics for the Horizontal Pod Autoscaler (HPA), allowing workloads to scale reliably without external monitoring dependencies. 🔗 Scaling on custom metrics in GKE

Google Cloud Next 2026 and onwards introduced many upgrades across the Data Analytics portfolio, shifting the focus toward the Agentic Data Cloud, an AI-native cross-cloud lakehouse, and real-time streaming operations. The announcements fundamentally redefine how data is governed, processed, and acted upon by AI agents.

Below is the summary of key announcements by service:

BigQuery & BigQueryML

• Fluid Scaling & Advanced Runtime Optimizations: Foundational architectural updates delivered massive performance boosts. Customers are seeing a 30% jump in overall query speed for large datasets (10TB) and a 62% improvement for small queries, while simultaneously improving cost-efficiency by 32%.

• Stateful Continuous Queries (Preview): Processing real-time data is now as easy as writing SQL query. Continuous queries now support Tumbling windows and stream-to-stream INNER JOINs, which decrease the friction from setting up realtime pipelines even more.

• BigQuery Graph Analytics (Preview): Bring all your data and calculations together in an overlay property graph and query it using GQL (ISO standard). Users can build graphs directly over existing relational data using simple CREATE PROPERTY GRAPH statements—ideal for fraud detection and chatbot grounding.

• Native GenAI, Semantic Functions & Vector Search: s: BQML natively integrates with Google's latest AI models directly via SQL, bringing support for TimesFM (time-series forecasting - GA), Zero-Shot Tabular Foundation Models (preview) , and ObjectRef (GA) for multimodal data handling. Additionally, the new fully-managed Vector Search provides autonomous embedding generation at scale using the AI.EMBED() function (GA).

• Data Agents (GA): BigQuery Studio is now an integrated GenAI-centric hub. The new Data Engineering Agent and Data Science Agent drastically reduce pipeline development time, while Conversational Analytics allows natural language exploration over multimodal data. To support multi-agent workflows, new Agent Analytics Plugins allow seamless integration to log detailed interactions from frameworks like ADK and LangGraph with a single line of code.

Lakehouse (formerly BigLake) & Managed Spark (formerly Dataproc)

• Read/Write Interoperability for Apache Iceberg (preview): The openness of Iceberg is now combined with BigQuery's enterprise-grade analytics. Customers can seamlessly read and write between BigQuery and open-source engines (like Spark, Flink, and Trino) using the new Iceberg REST Catalog, avoiding data silos and zero-copy federation across AWS, Azure, and others.

• Lightning Engine for Spark: The new Managed Service for Apache Spark (formerly Dataproc)  introduces the Lightning Engine. This fully Spark-compatible engine accelerates interactive and batch workloads, providing massive performance improvements over competitors.

Streaming & Pipelines: Pub/Sub, Dataflow & Managed Airflow (formerly Composer)

• Pub/Sub Subscriptions & Gemini Integrations: Single Message Transforms (SMT) for AI inference allow real-time sentiment analysis and entity extraction in-flight. Pub/Sub also added new native Bigtable Subscriptions. Coming soon, Gemini in Pub/Sub will offer intelligent schema inference, AI-assisted operations, and SMT code generation.

• Dataflow Job Builder & Lakehouse Integration: Dataflow job builder now supports external Iceberg REST Catalogs (IRC) as a source, enabling you to ingest data directly into Lakehouse for Apache Iceberg tables without writing code. Additionally, you can seamlessly register existing Apache Parquet files from cloud storage (e.g., Cloud Storage, Amazon S3) into Lakehouse Iceberg tables without moving or rewriting the underlying data.

• Dataflow Event-Driven Agents & GPU Support: Developers can use the Model Context Protocol (MCP) to define logic for real-time responses to data events. To accelerate these workloads, Dataflow now supports NVIDIA RTX Pro 6000 GPUs (accelerator type nvidia-rtx-pro-6000), which are highly recommended for large, medium, and small model inference on your Apache Beam pipelines.

• Managed Airflow 3.1 (formerly Composer): Decoupled architecture for a robust separation between the entire Airflow system and the execution layer for better scalability and enhanced security. Additionally Airflow 3.1 introduced real-time event-driven scheduling, native DAG versioning, and Agentic Troubleshooting tools in the Cloud Console to automatically assist in minimizing Mean Time To Recovery (MTTR) for failed pipelines. More information here.

Knowledge Catalog (formerly Dataplex)

• Always-on Universal Context Engine: Designed as the backbone for the Agentic Data Cloud, Knowledge Catalog governs your entire data estate across SaaS apps, databases, and analytical tools, helping agents execute complex tasks with accuracy.

• Key H1'26 Launches: Major automated cataloging features are rolling out, including support for the Lakehouse Iceberg Rest Catalog (GA) and AlloyDB (GA). Additional features include Semantic Search (GA), Data Products (GA), and a Business Glossary with at-scale imports and reusable data quality rules (GA). Unstructured data insights and Context Retrieval APIs are also available in Preview.

Looker & Data Studio (formerly Looker Studio)

• Conversational Analytics & Dashboard Agents: Empower dashboard viewers to go beyond static reports and chat directly with their dashboards to get tailored, natural language answers. It includes full Show Reasoning capabilities to guarantee transparency into how AI models generated the insights.

• Agentic LookML Development (Preview): A new LookML Agent integrated with a VS Code Extension empowers developers to easily write, edit, and validate LookML code, offering LLM flexibility to support creation based directly on design documents or mockups.

• New Consumption-Based Pricing Pilot: Looker introduced a highly predictable pricing model that pairs a fixed monthly subscription for baseline usage with flexible, usage-based overages (events & tokens) to scale as your business grows.

• Colab Data Apps (Preview): You can now convert Python visualizations and notebooks directly into interactive Colab Data Apps, making advanced data science outputs readily available to business users within Data Studio.

🤖 AI Security updates

• Model Armor for Agent Gateway GA: Google announced the General Availability of Model Armor inline integration for Agent Gateway, allowing security teams to apply content security templates and sanitization policies directly to agent-to-model and agent-to-tool traffic.
  🔗 Configure Model Armor on Agent Gateway

• Multimodal Image Screening in Model Armor: In preview, Model Armor now supports screening images within both prompts and model responses, extending runtime risk mitigation to protect visual modalities against adversarial injection.
  🔗 Model Armor Multimodal Image Screening

• Agent Gateway & Registry Policy Integration: Google released comprehensive documentation and design blueprints for governing agentic  with the Agent Registry to enforce strict tool-use boundaries.workloads with Agent Gateway, illustrating tight policy coupling
  🔗 Governing Agentic Workloads with Agent Gateway

🛡️ Security product updates

• Google SecOps June Release: Google Security Operations introduced advanced Time-To-Live (TTL) data retention settings in Google SecOps data tables, alongside streamlined parser operations and security webinars to enhance threat hunting.
  🔗 What's New in Google SecOps

• Wiz Runtime Sensor for Windows GA: Extending unified security across hybrid architectures, Wiz launched General Availability for the Windows Runtime Sensor, offering real-time memory-safe threat monitoring and DLL risk validation on Windows Server instances.
  🔗 Announcing the Wiz Runtime Sensor for Windows

• VPC-SC Perimeters for Agentic Workloads: Security teams can now leverage newly designed VPC Service Controls to prevent prompt-hijack-driven data exfiltration and unauthorized cloud-to-cloud data copying by misbehaving autonomous AI agents.
  🔗 Securing Agentic AI: What's New in VPC Service Controls

• IDC MarketScape SIEM Leadership: Google has been named a Leader in the 2026 IDC MarketScape SIEM evaluation, validating the frontend automation of Google Security Operations and the integration of Mandiant threat visibility.
  🔗 Google Named a Leader in SIEM 2026 Vendor Assessment

⚙️ Secure SDLC & Autonomous Posture Management

• Operationalizing AI Threat Defense Framework: Google Cloud CISO Chris Betz detailed a structured, four-step framework—Prepare, Scan/Prioritize, Remediate, and Monitor—built on findings that harness quality and compiler validation loops are more critical than raw model intelligence for auto-patching.
  🔗 Cloud CISO Perspectives: The 4 Lessons That Guided AI Threat Defense

🕵️ Frontline Threat Intelligence & Disruption

• NetNut Proxy Platform Takedown: In coordination with the FBI and Lumen, Google's Threat Intelligence Group successfully disrupted the NetNut residential proxy network (Popa botnet), cutting off 2 million infected consumer nodes.
  🔗 Dismantling Malicious Residential Proxy Networks

• Cisco Catalyst SD-WAN Zero-Day (CVE-2026-20245): Mandiant reported active zero-day exploitation of a high-severity command injection flaw, allowing threat actors who bypassed WAN authentication to escalate administrative permissions to root Unix access.
  🔗 Zero-Day Exploitation in Cisco Catalyst SD-WAN Manager

• ShinyHunters PeopleSoft RCE Attacks (CVE-2026-35273): Extortion actor UNC6240 targeted higher education and public sector networks, exploiting an unpatched SSRF-to-deserialization chain in Oracle PeopleSoft Environment Management Hub endpoints.
  🔗 ShinyHunters Exploitation of Oracle PeopleSoft

🔗 Direct link to the official Google developer blogs.

🚀 Google Antigravity (Suite, CLI, & SDK)

Google separated the platform hub into a standalone app, introducing a dedicated CLI and a separate IDE download.

Agentic Capabilities: Version 2.2.1 restored the IDE view button, tripled quotas, doubled content length, and added support for nested subagents and PDFs.

Antigravity SDK: The foundation for building custom automated workflows, enabling developers to build tools like agentic PR reviewers for enterprise codebases.

Enterprise Integration: Now supports official cloud credentials, terminal sandboxing, and an auto-included security extension.

🔗 Antigravity SDK Announcement

✨ Google AI Studio

Full-Stack Vibe Coding: Integrates backend infrastructure directly into the rapid-prototyping interface without requiring a credit card.

Android App Creation: Build and live-preview native, Kotlin-based apps directly in the browser from a single prompt via an embedded emulator.

Vibe Coding Mechanics: Seamlessly attach Cloud SQL, Firebase Auth, or Firestore databases to transition from a simple prompt to a fully functional app.

🔗 Build Native Android Apps in Google AI Studio

🧠 Gemma 4 & Gemini Nano

Gemma 4: An open multimodal framework optimized for reasoning. The 12B variant brings local, resource-optimized agentic workflows directly to laptops via Google AI Edge.

Gemini Nano: Designed for on-device use, running natively in Android's AICore to leverage hardware for low-latency inference.

🔗 Bringing Gemma 4 12B to your Laptop

The Gemini Enterprise Agent Platform has launched the General Availability of multiple pending orders for Provisioned Throughput, enabling organizations to queue up to seven model orders simultaneously. This allows enterprises to schedule capacity increases for predictable traffic spikes, such as promotional events, ensuring rapid response times for autonomous agent fleets. By automating throughput scaling, developers can eliminate manual adjustments and prevent quota exhaustion during peak demand periods. 

https://docs.cloud.google.com/gemini-enterprise-agent-platform/release-notes#July_01_2026

Google Cloud announced the General Availability of its Cloud Run Remote Model Context Protocol (MCP) server, offering a serverless, low-latency connection layer for AI agents via Streamable HTTP architecture. This allows developers to easily connect agents to databases and APIs without managing complex custom code or persistent WebSocket connections. Teams can deploy standard MCP-compatible containers, eliminating infrastructure boilerplate and enabling independent agent engines to share tool registries.

Google Cloud Next 2026 introduced substantial upgrades across its database portfolio, shifting the focus toward Agentic AI enablement, hybrid cloud flexibility, and automated AI operations. Below is the summary of announcements by service with the link to more info.
A direct link to the official Google Cloud announcement blogs.
What's New for Google Cloud Databases at Next '26

🚀 Database Center & AI Observability

Before diving into individual services, Google introduced a core platform-wide capability that affects most of these databases:

• Gemini-powered Database Observability Agent: A new proactive AI agent that continuously monitors performance and health across AlloyDB, Bigtable, Cloud SQL, and Spanner. It identifies the root causes of issues and delivers automated remediation playbooks to reduce operational downtime.

• 🔗 Database Center Improvements

🧠 Memorystore

• Move-to-Managed Migration Paths: Google introduced a streamlined, managed migration tool specifically designed to help users modernize self-managed Redis and Valkey instances by transitioning them smoothly into Memorystore.

• New Node Sizes & Core Modules: To provide better price-performance flexibility, Memorystore added smaller and larger node sizes alongside native support for the popular Valkey-bloom and Valkey-json modules.

• Enterprise-Grade Security: The platform added robust access security features including custom Access Control Lists (ACLs), token-based authentication, and Flexible Certificate Authorities.

• 🔗Memorystore for Valkey 9.0 GA Announcement

🛢️ Cloud SQL

• Move-to-Managed Experience (Preview): Embedded directly into the Cloud SQL console and Database Center, this update upgrades the Database Migration Service. For PostgreSQL, it introduces advanced capabilities like parallel consolidation of multiple source databases into a single live target, and support for migrating tables that lack primary keys.

• Security & Observability Integrations: Cloud SQL received native brute-force connection protection and air-gapped recovery options, alongside integration into the new fleet-wide Gemini slow query analysis tools.

• 🔗Cloud SQL Release Notes

⚡ AlloyDB

• AlloyDB AI-Powered Search at Scale (Preview): AlloyDB can now scale to handle up to 10 billion vectors using Google’s custom ScaNN index, delivering vector queries up to 6 times faster than standard PostgreSQL HNSW indexes. Native BM25 support is also coming soon to empower high-speed hybrid search engines.

• New and Optimized AI Functions (Preview): Google added ai.analyze_sentiment and ai.summarize into AlloyDB's built-in suite of LLM functions. Additionally, the existing ai.if fraud-detection function received a highly optimized mode to boost performance and slash processing costs.

• 🔗AlloyDB Release Notes

🌐 Spanner

• Spanner Omni: This milestone release allows users to run Spanner’s globally consistent, multi-model database environment entirely outside of Google Cloud—including on-premises, across alternative cloud environments, or locally on developer laptops.

• Spanner Columnar Engine (GA): This engine accelerates analytical queries on live operational data, executing data scans up to 200 times faster. It now natively supports Iceberg tables, continuous reverse ETL from BigQuery, accelerated federated queries.

• 🔗 Spanner Release Notes

📊 Bigtable

• Enterprise Plus Edition & Hybrid Storage: Bigtable introduced a brand-new Enterprise Plus edition featuring nodes with a hybrid storage architecture spanning RAM, SSD, and HDD. This allows the database to intelligently keep frequently accessed cache data in-memory while delegating older records to cost-effective archival storage under a single service.

• 🔗Bigtable Release Notes

🔥 Firestore

• Third-Party AI Agent Skills & MCP (GA): Connecting Firestore to external programming agents like Claude Code, Cursor, and Codex is now natively streamlined via the General Availability of Firestore Skills and the Firestore remote Model Context Protocol (MCP) service.

• Natural Language to MongoDB Queries (Preview): Powered by Gemini Code Assist in the Google Cloud console, developers can type natural language questions to automatically generate complex, MongoDB-compatible Firestore queries.

• 🔗 Official Blog: Firestore: Agentic AI, Search, and MongoDB Compatibility