Product mappings

Use the following list to identify the appropriate Google Cloud products and features for each component in the confirmed technical decomposition:

Table of Contents

Cloud Run networking

• Recommended primary configuration: Regional External Application Load Balancer combined with Cloud Armor for HTTP/HTTPS/WebSocket ingress, and Direct VPC egress for Cloud Run private network access.
• Alternative product 1: Global External Application Load Balancer
  • Pros: Single anycast IP, global IPv6 termination, and low-latency routes to globally distributed backend services.
  • Cons: Terminates TLS globally at edge locations, which might not comply with strict regional data residency regulations.
• Alternative product 2: Internal Application Load Balancer
  • Pros: Securely exposes Cloud Run services internally within the VPC to meet internal ingress criteria, terminates TLS with trusted certificates, and supports Cloud Armor backend security policies.
  • Cons: Requires that you configure serverless network endpoint groups (NEGs) as backends and manage load balancer resources.
• Alternative product 3: Private Service Connect interface
  • Pros: Secure private VPC connections for Gemini Enterprise Agent Runtime that uses network attachments.
  • Cons: Limited to RFC 1918 routable subnet ranges, requires proxy setup for non-routable/internet destinations.

Frontend

• Recommended primary product: Cloud Run
• Alternative product 1: Firebase App Hosting
  • Pros: Automated builds and deployment pipeline from GitHub, optimized for modern framework integrations.
  • Cons: Less control over container configurations, limits customization of low-level networking.
• Alternative product 2: Google Kubernetes Engine (GKE)
  • Pros: Maximum control over routing, scaling, and custom container runtimes.
  • Cons: Significant infrastructure management complexity and cost overhead.

Agent development framework

• Recommended primary product: Agent Development Kit (ADK).

Agent-to-agent communication

• Recommended primary product: Agent Gateway for governed Agent2Agent (A2A) connectivity.
  • Pros: Provides client-to-agent ingress, agent-to-anywhere egress, and natively integrates IAM, Model Armor, Agent Registry, and Agent Identity.

Agent runtime

• Recommended primary product: Gemini Enterprise Agent Runtime
  • Pros: Fully managed runtime that supports built-in memory/sessions (Agent Platform Sessions), secure code execution sandbox, and native connection to remote MCP servers (like those hosted on Cloud Run).
  • Cons: Limited to the supported languages (see the supported Language list) and doesn't host custom Model Context Protocol (MCP) servers directly (they must be hosted externally on Cloud Run or GKE).
• Alternative product 1: Cloud Run
  • Pros: Serves custom container instances scaling to zero; highly performant when hosting custom language engines, local API layers, or custom MCP servers.
  • Cons: Requires managing container builds and continuous delivery configurations, can't host or serve Gemini models.
• Alternative product 2: Google Kubernetes Engine (GKE)
  • Pros: Maximum infrastructure control, stateful pods, custom scaling.
  • Cons: High operational complexity and overhead.

Agent registry

• Recommended primary product: Agent Registry
  • Pros: Provides a unified catalog for agents, tools, and MCP servers. Optimizes routing in multi-agent (A2A) paths.

Model runtime

• Recommended primary product: Gemini Enterprise Agent Platform
• Alternative product 1: Cloud Run
  • Pros: Serverless hosting for containerized open/custom models like Gemma. Can be configured to autoscale.
  • Cons: Cannot serve Google Gemini models.
• Alternative product 2: Google Kubernetes Engine (GKE)
  • Pros: Maximum control over inference server on GPU/TPU nodes, cheap for predictable high volume.
  • Cons: Cannot run Google Gemini models, high cluster management overhead.

Model selection

• Recommended primary product (Text): Gemini Flash
• Recommended primary product (Audio/Video): Gemini Flash with Gemini Live API
• Alternative product 1: Gemini Pro
  • Pros: Highest capability for reasoning, complex instructions, context tracking, and multi-agent coordination.
  • Cons: Higher request cost and latency, which makes it less suitable for real-time conversational requirements.

Model input and output inspection

For security, always include model input and output inspection.

• Recommended primary product: Model Armor

VPC connection to databases

An agent sends queries through this connector to securely access resources in the Virtual Private Cloud (VPC) network used for storage resources in this architecture.

• Recommended primary product: Serverless VPC Access connector
• Alternative product 1: Direct VPC egress
  • Pros: Lower latency, lower resource cost, and avoids throughput scaling bottlenecks.
  • Cons: Requires specific routing and subnet configurations.

Agent memory

Short-term memory

• Recommended primary product: Agent Platform Sessions
  • Pros: Native, built-in managed memory for Gemini Enterprise Agent Runtime, low-ops, the default choice for most cases since it eliminates the need to provision and manage external databases.
  • Cons: Requires an Agent Runtime instance and is limited to the managed Agent Platform Sessions service (regional availability applies), although the agent itself can still run on Cloud Run or GKE.
• Alternative product 1: Memorystore for Redis Cluster
  • Pros: High performance, sub-millisecond read/write latency. Ideal escalation path when you require microsecond response times and high throughput.
  • Cons: Requires managing Redis nodes, VPC routing/peering configuration, and has a higher base cost.
• Alternative product 2: Firestore
  • Pros: Fully managed NoSQL database with serverless scaling, excellent for persistent state storage.
  • Cons: Higher latency compared to in-memory stores, which might affect real-time interactive applications.
• Alternative product 3: Cloud SQL (with ADK DatabaseSessionService)
  • Pros: Strong relational guarantees and ACID compliance for complex querying over session state.
  • Cons: Higher connection overhead and latency, and requires managing database instances.

Long-term memory

• Recommended primary product: Memory Bank
• Alternative product 1: Firestore
  • Pros: Serverless, highly scalable document database that is ideal for storing large volumes of persistent knowledge.
  • Cons: Requires custom implementation for semantic search, embedding, and retrieval logic compared to purpose-built memory solutions.

Database (for search and retrieval)

• Recommended primary product: Google Cloud Databases Use the recommendations listed on this page to help the user choose the appropriate database option (e.g., Cloud SQL, AlloyDB).
• Alternative product 1: Compute Engine (for self-hosted databases)
  • Pros: Full control over database configurations, OS accessibility, and custom database engines or extensions.
  • Cons: High operational overhead to manually manage backups, patching, scaling, and high availability.

Agent tools

• API Management Platform (Enterprise Scale):
  • Recommended product: Apigee API hub.
  • Use case: Best for managing, securing, and monitoring a large number of API-based tools at an enterprise scale. It allows agents to connect to data instantly through prebuilt connectors or custom APIs.
• Model Context Protocol (MCP):
  • Recommended primary products:
    • Google Cloud MCP servers (fully managed by Google for connecting to Google Cloud services).
    • Cloud Run or GKE (for deploying custom, self-hosted MCP servers as containerized applications).
    • Use case: Best for building modular or multi-agent systems that require interoperable and reusable tools, decoupling the agent's logic from specific tool implementations.
• Built-in Tools:
  • Recommended primary framework: Agent Development Kit (ADK).
  • Use case: Best for common tasks (e.g., web search, code execution in a secure environment) to accelerate initial development without configuring external communication protocols.
• Custom Function Tools:
  • Use case: Best for integrating with specific internal or third-party APIs that do not have an MCP server, or for exposing an "agent-as-a-tool" function in multi-agent orchestration.