GCP Release Notes: April 22, 2026

Agent Registry

Issue

The following known issues affect Agent Registry:

• Search location filtering: When calling the SearchAgents or SearchMcpServers APIs for the global location, the results might incorrectly include resources from us and eu multi-regions.
• URN mismatch: When searching for agents or MCP servers in the Google Cloud console, the page might display an invalid URN format in the search results list.
• Console error: Users who actively switch between tabs on the MCP server details page in the Google Cloud console might encounter an unexpected throttling error.

Announcement

The Agent Registry remote Model Context Protocol (MCP) server is available in Preview. You can connect your AI applications to the Agent Registry MCP server to dynamically discover other agents, endpoints, and MCP servers available in your environment.

For more information, see Use the Agent Registry remote MCP server.

Announcement

Agent Registry is available in Preview. Agent Registry is a centralized catalog for discovering and registering agents and Model Context Protocol (MCP) servers.

For more information, see the Agent Registry overview.

AlloyDB for PostgreSQL

Feature

Database Migration Service quick-start migrations (in Preview) are now integrated into AlloyDB to provide a lightweight, continuous migration flow. This feature automates setup for sources with private IPs in a VPC network, including Cloud SQL for PostgreSQL and self-managed databases on Compute Engine.

For more information, see Quick-start migrations overview in the Database Migration Service documentation.

Application Design Center

Feature

Application Design Center supports the following components in General Availability:

• CA Service
• Private Service Connect Endpoint
• Private Service Connect Producer

Application Design Center supports the following components in Preview:

• Authorization Extension
• Authorization Policy
• Authorization Policy Extension
• Compute Address
• Firebase AI Logic Prompt Template
• Firebase Multi-Platform App
• Routes
• Agent Registry Agent
• Agent Registry Binding
• Agent Registry Endpoint
• Agent Registry MCP Server
• Firestore Security Rules
• IAM Connector
• Model Armor Floor Setting
• Model Armor Template
• VPC Network
• Cloud Workflows
• Firebase AI Logic
• Firebase App Check
• Firebase Authentication
• Compute Firewall
• Cloud KMS
• Internal Load Balancer
• Agent Registry Service
• Artifact Registry
• Cloud Run functions
• Cloud Tasks
• Managed Airflow
• Cloud DNS Managed Zone
• Cloud DNS Response Policy
• Document AI
• Cloud NAT
• Cloud Router
• Cloud Router Interface
• Secure Web Proxy
• Compute Instance
• Cloud Scheduler
• Agent Platform Runtime

Feature

If your application deployment fails, you can troubleshoot and fix errors (Preview). For more information, see Troubleshoot and fix deployment issues.

Feature

Create a composite template (Preview) using multiple application templates and components. For more information, see Design composite templates.

Feature

You can store templates and applications in the following regions:

• Tokyo, Japan (asia-northeast1)
• Seoul, South Korea (asia-northeast3)
• Taiwan (asia-east1)
• Hong Kong (asia-east2)
• Delhi, India (asia-south2)
• Singapore (asia-southeast1)
• Jakarta, Indonesia (asia-southeast2)
• Melbourne, Australia (australia-southeast2)
• Hamina, Finland (europe-north1)
• Stockholm, Sweden (europe-north2)
• Warsaw, Poland (europe-central2)
• St. Ghislain, Belgium (europe-west1)
• London, England (europe-west2)
• Frankfurt, Germany (europe-west3)
• Milan, Italy (europe-west8)
• Paris, France (europe-west9)
• Turin, Italy (europe-west12)
• Eemshaven, Netherlands (europe-west4)
• Zurich, Switzerland (europe-west6)
• Madrid, Spain (europe-southwest1)
• Columbus, Ohio (us-east5)
• Ashburn, Virginia (us-east4)
• The Dalles, Oregon (us-west1)
• Los Angeles, California (us-west2)
• Salt Lake City, Utah (us-west3)
• Las Vegas, Nevada (us-west4)
• Council Bluffs, Iowa (us-central1)
• Dallas, Texas (us-south1)
• Montréal, Canada (northamerica-northeast1)
• Toronto, Canada (northamerica-northeast2)
• Queretaro, Mexico (northamerica-south1)
• São Paulo, Brazil (southamerica-east1)
• Santiago, Chile (southamerica-west1)
• Johannesburg, South Africa (africa-south1)
• Doha, Qatar (me-central1)
• Tel Aviv, Israel (me-west1)
• Global

For more information, see the following:

• A list of available App Design Center locations.
• To select and manage a region, see Select a region.
• To share a catalog between spaces, see share a catalog.

Feature

Use the following Google-provided application templates:

• Single region GKE cluster and workload (Preview)
• Enterprise-grade production GKE cluster and workload (Preview)
• AI Pre-trained Inference GKE cluster and workload (Preview)

BigQuery

Feature

You can now use the visual graph modeler in BigQuery Studio to define BigQuery graph nodes and edges from your BigQuery tables and edit graph schema. This feature is available in Preview.

Announcement

Dataproc is now called Managed Service for Apache Spark. The names for associated API, client library, CLI, and Identity and Access Management (IAM) resources remain unchanged.

Announcement

BigLake is now called Google Cloud Lakehouse. BigLake metastore is now called the Lakehouse runtime catalog. The names for associated APIs, client libraries, CLI commands, and Identity and Access Management (IAM) remain unchanged and still reference BigLake.

Announcement

Dataplex Universal Catalog is now called Knowledge Catalog. The API, client library, CLI, and Identity and Access Management (IAM) names remain unchanged. For more information, see Knowledge Catalog overview.

Announcement

Looker Studio is now called Data Studio. The website and endpoint change from lookerstudio.google.com to datastudio.google.com. You do not need to update your reports for this change, as Data Studio automatically redirects to the new domain. However, if your company uses proxies to restrict access to external sites, your IT administrator needs to add the new domain to your access control list (ACL). The names for associated API, client library, CLI, and Identity and Access Management (IAM) resources remain unchanged. For more information, see Data Studio returns as new home for Data Cloud assets.

Feature

BigQuery graphs now support the following features:

• You can query graphs using natural language in Conversational Analytics.
• You can add descriptions and synonyms to the labels and properties in your graphs.
• For some types of graphs you can define measures, which lock an aggregation to a key to help you perform complex aggregations without overcounting. To query measures, you transform your graph into a flattened table by using the GRAPH_EXPAND TVF, and then query measures in that table with the AGG function.

These features are in Preview.

Feature

You can now use the Data Engineering Agent to build, modify, and troubleshoot data pipelines in BigQuery. This feature is generally available (GA).

Bigtable

Feature

The Bigtable editions feature is generally available (GA). Bigtable editions introduces advanced features in performance, analytic query capability, and resource management. You can choose between the Enterprise and Enterprise Plus edition to select the right capabilities for your workloads. For more information, see Editions overview.

Feature

Bigtable provides an in-memory tier as part of its hybrid storage architecture. This tier provides sub-millisecond read latency and high throughput for time-sensitive data with independent vertical scaling to handle traffic surges. The in-memory tier is available only in the Enterprise Plus edition in Preview. For more information, see In-memory tier overview.

Feature

Bigtable tiered storage limit increases from 32 TB to 64 TB per node. This expansion provides higher storage density to support retention of larger volumes of infrequently accessed data and is available only in the Enterprise Plus edition. Tiered storage is available in Preview.

Feature

As part of Bigtable editions, you can use Data Boost to read data from tiered storage and HDD clusters. This feature is available only in the Enterprise Plus edition and is generally available (GA).

Feature

As part of Bigtable editions, you can use Data Boost to run GoogleSQL queries. This feature is available only in the Enterprise Plus edition and is generally available (GA). For more information, see High-throughput SQL analysis with Data Boost.

Feature

As part of Bigtable editions, you can configure which cluster in a replicated instance is used for automated backups. This feature provides greater cost control and backup resource management. This feature is available only in the Enterprise Plus edition and is generally available (GA). For more information, see Bigtable backups overview.

Cloud Composer

Feature

Managed Service for Apache Airflow supports Gemini Cloud Assist Investigations capabilities. The new troubleshooting agent can now troubleshoot failed Airflow task instances and DAG runs. The feature is available through Gemini Cloud Assist Investigations, which is currently accessible in Private Preview.

Cloud Database Migration Service

Feature

You can now use quick-start migrations for homogeneous PostgreSQL migrations to Cloud SQL for PostgreSQL and AlloyDB for PostgreSQL.

Quick-start migrations are a lightweight continuous migration flow where Database Migration Service can automatically set up everything you need to migrate sources that have a private IP assigned in a VPC network, such as self-managed databases on Compute Engine or Cloud SQL for PostgreSQL instances. This feature is in Preview.

Cloud Hub

Feature

Cost optimization with Gemini Cloud Assist provides the following additional features:

• In the Gemini Cloud Assist chat panel, get an explanation for cost changes for supported resources.
• On the Optimization page in Cloud Hub, get insights about recent cost changes and related changes in resource usage.

For more information, see Optimize costs with Gemini assistance.

Feature

App Topology is in Preview. App Topology lets you query data about your resources and applications from multiple sources, and then view the correlated data as a topology graph.

Cloud Load Balancing

Feature

Policy profiles in authorization policies let you define the type of authorization being performed at the load balancer. This feature is available in Preview.

You can choose from the following profile types:

• Request authorization profile (REQUEST_AUTHZ): Evaluates access based on HTTP request headers. Authorization decisions can be made directly or delegated to custom services. This is the default profile.

• Content authorization profile (CONTENT_AUTHZ): Enables deep inspection of application payloads (headers, body, and trailers). This is used for content-based security, such as blocking prompt injection attacks and preventing sensitive data leaks. Authorization decisions are always delegated.

Policy profiles are supported for the following Google Cloud services:

• Regional external Application Load Balancers
• Regional internal Application Load Balancers
• Agent Gateway (Preview)
• Secure Web Proxy

To learn more about policy profiles, see Authorization policy overview.

Cloud Monitoring

Feature

Application Monitoring in Google Cloud provides both agent observability and application observability. Your Application Monitoring dashboards display performance metrics, including the error rates and token usage of your AI resources. Those metrics can help you understand the health and performance of your AI resources.

To learn more, see the following:

• Agent observability
• Application Monitoring overview
• Investigate applications, services, and workloads
• View AI resources

Cloud Number Registry

Announcement

Preview: Cloud Number Registry provides IP address management (IPAM) capabilities to let you view, manage, and plan your IP address usage in Google Cloud.

Cloud SQL for PostgreSQL

Feature

Database Migration Service quick-start migrations are now integrated into Cloud SQL for PostgreSQL to provide a lightweight, continuous migration flow. This feature automates setup for sources with private IPs in a VPC network, including Cloud SQL for PostgreSQL instances and self-managed databases on Compute Engine.

For more information, see Quick-start migrations overview in the Database Migration Service documentation. This feature is in Preview.

Dataflow

Feature

Dataflow job builder now supports external Iceberg REST Catalogs as a source. You can now ingest data from external Apache Iceberg REST catalogs (IRC) directly into Google Cloud Lakehouse tables using Dataflow’s job builder UI without writing code. For more information, see Import data from external Iceberg catalogs to Lakehouse using Dataflow.

Eventarc

Feature

Eventarc support for creating triggers for direct events from Gemini Cloud Assist is available in Preview.

Gemini Enterprise Agent Platform

Change

Vertex AI to Gemini Enterprise Agent Platform naming changes

The table below lists all of the features that have been transitioned from Vertex AI and what their new names are in Agent Platform.

Change

Initial release of Gemini Enterprise Agent Platform

This initial release includes (but is not limited to) the following releases or changes:

• Change Vertex AI is now part of Gemini Enterprise Agent Platform. Information on model support for Vertex AI is now under Gemini Enterprise Agent Platform > Models.
• Change Agent Builder is now part of Gemini Enterprise Agent Platform. Features have been renamed as follows:
  • Agent Engine is now Agent Runtime.
  • Agent Builder Sessions is Agent Platform Sessions.
  • Memory Bank is now Agent Platform Memory Bank.
• Change Agent Runtime now supports long-running operations (up to 7 days).
• Change Agent Runtime now supports sub-second cold starts.
• Change Provisioning for Agent Runtime has been reduced to less than 1 minute.
• Release You can now use your own custom-built containers when you deploy agents with Agent Runtime.
• Change When creating a Session, you can specify your own session ID.
• Release Memory Bank now enables continuous event streaming with automated memory generation triggered by configurable criteria like event count or idle time. For more information, see Ingest events.
• Release Memory Bank now automatically maintains an immutable version history of memories through revision resources. For more information, see Memory revisions.
• Release Agent Identity for General Availability. Agent Identity helps let your agent securely authenticate to MCP servers, cloud resources, endpoints, and other agents, either acting as itself or acting on behalf of the end user.
• Release Agent Gateway for Private Preview. Agent Gateway is the networking component of the Gemini Enterprise Agent Platform ecosystem. It secures and governs connectivity for all agentic interactions, whether they occur between users and agents, agents and tools, or among agents themselves.
• Release Agent Registry for Public Preview. Agent Registry is a centralized, unified catalog that lets you store, discover, and govern Model Context Protocol (MCP) servers, tools, and AI agents within Google Cloud.
• Release New IAM governance policies are available in Private Preview.
• Release Agent Observability for Preview. Agent Observability in Gemini Enterprise Agent Platform provides comprehensive visibility into the performance, behavior, and health of your deployed agents and Model Context Protocol (MCP) servers. By monitoring key metrics, tracing execution paths, and observing your multi-agent system as a whole, you can diagnose issues, optimize resource consumption, and improve the reliability of your agents.
• Release Gemini Embedding 2 (gemini-embedding-2) for General Availability.
• Release The Gemini Deep Research Agent, a pre-built agent designed to help you plan, execute, and synthesize multi-step research tasks. It uses Gemini 3.1 Pro to bridge the gap between public web data and private enterprise context by simultaneously grounding research across three distinct, high-fidelity data streams.
• Change Google Cloud console navigation: The navigation menus under Agent Platform (formerly Vertex AI) and Data Analytics have been updated to centralize agentic products and features. Bookmarked links will continue to work via automatic redirects.

Google Distributed Cloud (software only) for VMware

Announcement

Google Distributed Cloud (software only) for VMware 1.33.700-gke.71 is now available for download. To upgrade, see Upgrade clusters. Google Distributed Cloud 1.33.700-gke.71 runs on Kubernetes v1.33.5-gke.2200.

If you are using a third-party storage vendor, check the Google Distributed Cloud-ready storage partners document to make sure the storage vendor has already passed the qualification for this release.

After a release, it takes approximately 7 to 14 days for the version to become available for use with GKE On-Prem API clients: the Google Cloud console, the gcloud CLI, and Terraform.

Fixed

The following issues were fixed in 1.33.700-gke.71:

• Fixed vulnerabilities listed in Vulnerability fixes.
• Resolved an issue that caused VMware cluster upgrades from non-advanced clusters to advanced clusters to get stuck. The system attempted to update immutable fields in the Hub membership. With this fix, the cluster operator preserves the original membership fields during the upgrade process instead of attempting to overwrite them so that the migration to an advanced cluster completes successfully.
• Fixed an issue in advanced user clusters where the cloud.google.com/gke-nodepool label for workload node pools unexpectedly included an -np suffix. This caused pods using nodeSelector targeting the original pool name (such as Apigee workloads) to fail to schedule. For clusters on older versions experiencing this issue, you can manually set the expected label in the node pool configuration.
• Fixed an issue where setting the deprecated stackdriver.enableVPC field to true in a cluster configuration file would block upgrades to an Advanced Cluster. The stackdriver.enableVPC field has been deprecated and its setting is now ignored during the upgrade validation process. For clusters on older versions experiencing this issue, remove the field or set it to false in your configuration file before you upgrade.
• Fixed an issue where the node-problem-detector was incorrectly deployed onto non-Advanced VMware clusters. This caused the containerd runtime to continuously restart on affected nodes due to incompatible health check configurations, leading to ETCD/CRI failures (such as errors connecting to /run/containerd/containerd.sock) and unsuccessful cluster upgrades.

Feature

The following feature was added in 1.33.700-gke.71:

• Improved the resilience of migration from regular clusters to advanced clusters. If a migration attempt fails or is interrupted, you can now safely retry the process without manual cleanup. The system automatically reuses existing resources and the temporary bootstrap cluster when you retry. Important: If a migration fails, do not delete the bootstrap cluster, because deleting the bootstrap cluster can cause data loss and prevent recovery.

Google Distributed Cloud (software only) for bare metal

Announcement

Google Distributed Cloud (software only) for bare metal 1.33.700-71 is now available for download. To upgrade, see Upgrade clusters. Google Distributed Cloud for bare metal 1.33.700-71 runs on Kubernetes v1.33.5-gke.2200.

After a release, it takes approximately 7 to 14 days for the version to become available for installations or upgrades with the GKE On-Prem API clients: the Google Cloud console, the gcloud CLI, and Terraform.

If you use a third-party storage vendor, check the Google Distributed Cloud-ready storage partners document to make sure the storage vendor has already passed the qualification for this release of Google Distributed Cloud for bare metal.

Announcement

The following features were added in 1.33.700-gke.71:

A health check was added to detect when secrets or config maps mounted in pods become “stale,” or out-of-sync with the Kubernetes API server. This feature addresses scenarios where the Kubelet’s local cache fails to update with the latest versions of configuration data. The check performs the following actions:

• Iterates through all running pods on the node to verify their mounts.
• Compares the local data in the Kubelet’s atomic update symlink structure against the live objects and update timestamps in the API server.
• Uses a 5-minute threshold to prevent false positives caused by normal propagation delays. A mismatch is only reported as an error if the staleness persists for more than 5 minutes.

Fixed

The following issues were fixed in 1.33.700-gke.71:

• Fixed vulnerabilities listed in Vulnerability fixes.
• Fixed an issue where concurrent tasks on the same node failed when containerd restarts. After the fix, tasks are locked and run sequentially to ensure each task completes successfully before the next begins. Each lock is held for up to 20 minutes or until the task reaches success or failure. To bypass this safety mechanismrun and run tasks concurrently, add the following annotation to your cluster: baremetal.cluster.gke.io/concurrent-machine-update: "true".
• Fixed an issue where, during the machine initialization phase, the etcd-events pod read the stale data directory when it started and attempted to reuse the old member ID to rejoin the cluster instead of the new one. Trying to use the old member ID to rejoin the cluster resulted in an infinite retry loop and caused the cluster to reject the connection. The fix ensures that the system clears the /var/lib/etcd-events directory upon failure, and adds retry logic to kubeadm-reset to improve resiliency against transient API errors.
• Fixed an issue where node upgrades could hang indefinitely and bypass the 20-minute maintenance timeout. This issue occurred when a node contained completed pods within a namespace that was in a Terminating state. Because the Kubernetes Eviction API rejects operations in terminating namespaces, the cluster controller entered an infinite retry loop. The fix updates the drain process to skip eviction for pods in terminal phases, allowing the upgrade to proceed normally.
• Fixed an issue where Metrics API operations—including kubectl top, Horizontal Pod Autoscaling, and Vertical Pod Autoscaling could fail with TLS verification errors during certificate authority rotation. This occurred because the leaf certificate was not immediately renewed when the certificate authority was rotated, causing a temporary mismatch between the trusted certificate authority bundle and the certificate presented by the metrics server.

Identity and Access Management

Feature

Privileged Access Manager supports agent identities as grant requesters and approvers.

This feature is available in preview.

For more information, see Privileged Access Manager overview.

Feature

Agent Identity auth manager is available in preview. You can use Agent Identity auth manager to help securely authenticate your agents to third-party services using 3-legged OAuth, 2-legged OAuth, or API keys.

For more information, see Agent Identity auth manager.

Feature

Agent Identity is generally available (GA). Agent Identity provides a strongly attested, cryptographic identity for each agent that is tied to the lifecycle of the resource hosting the agent.

For more information, see Agent Identity overview.

Security Command Center

Feature

When Security Command Center is activated at the project level only, you can enable Vulnerability Assessment for Google Cloud on the single project.

Feature

Security Command Center has new predefined rules and controls:

• Additional predefined security graph rules to support Agent Runtime

• Additional support in existing correlated threats rules for Agent Runtime

• Additional runtime detectors in Agent Platform Threat Detection

• Additional Event Threat Detection rules to support AI agents

Feature

Security Command Center findings that are related to AI security risks are available in the Security tab of the Gemini Enterprise Agent Platform. The feature helps provide comprehensive visibility into findings, active threats, and attack path simulations. This feature requires Security Command Center Premium or Enterprise.

For more information, see View security findings.

VPC Service Controls

Feature

Preview stage support for the following integration:

• App Topology

Feature

Preview stage support for the following integration:

• Agent Registry

Vertex AI Search

Feature

Agent Search: MCP server (GA)

Agent Search has a Model Context Protocol (MCP) server hosted at the following endpoint: https://discoveryengine.googleapis.com/mcp

This feature is generally available (GA). For more information, see MCP Reference: discoveryengine.googleapis.com.

Feature

Agent Search: Dense reciprocal rank for custom ranking

You can use the dense reciprocal rank transformation function, drr, to customize search result ranking. It’s an improvement on the reciprocal rank function, rr. Using the dense reciprocal rank function leads to higher quality ranking when there are duplicate signal values.

Duplicate signal values are more common when the ranking formula contains the following types of signal:

• The boosting_factor signal
• The geo_distance() function signal
• Categorical and integer custom signals

This feature is generally available (GA). For more information, see Customize search results ranking.

Feature

Agent Search: Geodistance function for custom ranking (GA)

The geo_distance function can be used in custom ranking formulas to calculate the distance between a source location and a destination location. The function supports query locations extracted from natural language, explicitly provided coordinates, and addresses.

This feature is generally available (GA). For more information, see Custom ranking: Geodistance—a derived signal.

Feature

Agent Search: Filter searches by document-level relevance (GA)

When searching in your Agent Search app, you can specify document-level relevance filters so that only the documents that meet the filter threshold are returned as results.

You can specify either the relevance threshold or semantic-relevance threshold to filter documents by relevance based on keyword and semantic search similarity.

This feature is Generally Available (GA). For more information, see Filter searches by document-level relevance.

Change

Agent Search: Renamed from Vertex AI Search

The Vertex AI Search product has been renamed as Agent Search in the following contexts:

• The documentation set. See What is Agent Search?

What has not changed:

• The user interface in the Google Cloud console is still referred to as Vertex AI Search and AI Applications. See Vertex AI Search.

• The APIs still use the Discovery Engine API endpoints. See APIs and reference.

Despite the rebrand, the product functionality remains the same.

Source: Google Cloud Platform