Runtime vs Google Gemini Enterprise Agent Platform

Gemini Enterprise Agent Platform and Runtime at a glance

Gemini Enterprise Agent Platform is Google Cloud's platform for building and running AI agents. Google announced it at Cloud Next on April 22, 2026, as the evolution of Vertex AI, and says all Vertex AI services will be delivered through it going forward. It is organized around four jobs. Build: the open-source Agent Development Kit (ADK), the low-code Agent Studio, Agent Garden templates, and Model Garden with more than 200 models, including Gemini and Anthropic's Claude. Scale: Agent Runtime, with fast cold starts and long-running agents, plus Agent Sandbox and Memory Bank. Govern: Agent Identity, Agent Registry, and Agent Gateway with Model Armor. Optimize: simulation, evaluation, observability, and an optimizer. It is built for engineering teams that want to assemble agents on Google Cloud.

Runtime is the AI agent harness for payment and fintech teams: an operating system for building and running many agents across the org. Anyone in payment ops, risk, compliance, finance, underwriting, onboarding, or support builds agents from their SOPs. Agents work on their own isolated computers, reach your ledger, processor, and bank portals through APIs, databases, MCP servers, and a browser, and stop for a person before anything moves money. It is built for operations leaders who want agents working real queues this quarter.

Good products, different jobs. One is a set of excellent building blocks. The other is the finished harness.

How they differ

Building blocks vs a finished harness

Google gives engineers the parts: a framework (ADK), a managed runtime, memory, identity, a gateway for tool calls, and evaluation. They are well designed and they work together. What you build with them, and how ops teams use it, is up to your engineers.

Runtime ships those parts already assembled for operations work. The first agent is easy on any platform. Running agents on payment data needs isolated computers, scoped credentials, RBAC, approvals, an audit trail, evals, model routing, fallbacks, and a way for non-engineers to build and call agents from Slack or Teams. An engineering team can spend two quarters on that before the first agent touches real data. Runtime gives it to you on day one.

Where it runs and who holds the data

Google runs Agent Runtime, Memory Bank, and the governance services in Google Cloud. ADK itself is open source and can be containerized and run anywhere, but the managed platform is a Google Cloud service.

Runtime runs agent computers in your AWS, GCP, or Azure account, or fully self-hosted via Helm, and lets you bring your own sandbox provider. If you already run on Google Cloud, Runtime can run there too, so this is not a choice between Runtime and your cloud. If you run on more than one cloud, one harness covers all of them.

Models and harnesses

Google offers broad model choice through Model Garden, including Gemini, Gemma, and Claude, and ADK has adapters for other providers and locally running models.

Runtime routes across harnesses (Claude Code, Codex, OpenCode) and across models, with fallbacks when a provider goes down. You can serve open-weight models in your own cloud for PCI and PII work. Card numbers and SSNs are stripped from prompts and logs either way.

Guardrails designed for money movement

Google provides strong general governance: cryptographic agent identities, gateway policies on tool traffic, Model Armor against prompt injection and data leakage, and threat detection. Deciding which actions need human sign-off, and building that approval flow, is your team's design work.

Runtime starts agents read-only and requires approval before anything moves money: releasing a payout, applying a reserve, booking a ledger entry, replying to a sponsor bank. Credentials are masked and every run is recorded end to end, from the trigger through every query, tool call, approval, cost, and result. The record your sponsor bank, auditor, or examiner asks for stays with you, ready to export.

How you get to production

Google is self-serve for developers, with Google Cloud sales and partners for larger deployments.

Runtime pairs you with a forward-deployed AI engineer. The team built payment and fintech infrastructure at Hulu's payments team at Disney, Finix, Modern Treasury, and BlackRock's AI quant group. The FDE maps your processes, builds the first agents with your team, sets up guardrails, and trains admins. Your people keep using the software after the FDE leaves. When a process is solved, agents can turn it into a deterministic script that lives in your repos, so engineers own it if it becomes mission-critical.

Where Gemini Enterprise Agent Platform is stronger

• Depth of primitives. Agent Runtime, Memory Bank, Agent Identity, Agent Registry, and Agent Gateway are mature building blocks for a platform team that wants full control.
• Model catalog. More than 200 models in Model Garden, with first-party access to Gemini.
• Google Cloud integration. Batch and event-driven agents over BigQuery and Pub/Sub, and security tooling tied into Security Command Center.
• Breadth of use cases. It is horizontal: support, healthcare, retail, and internal knowledge agents, not only payments.
• Developer tooling. An open-source framework in five languages, simulation, continuous evaluation, and an optimizer for agent instructions.

Pricing

Gemini Enterprise Agent Platform is usage-based. Agent Runtime and sandbox environments are billed on compute at $0.085 per vCPU-hour plus memory per GiB-hour, and Google says idle time between turns is not billed for Runtime. Memory Bank, Sessions, and Gateway usage are metered, and model tokens are priced by model. The engineering time to build approvals, audit export, and ops-facing interfaces is not on the price list.

Runtime has public tiers: Free ($0, one session), Teams from $99 per seat per month, and Enterprise with custom pricing and self-hosting. The value to weigh it against is the work of the analysts you were about to hire.

Which should you choose

Choose Gemini Enterprise Agent Platform if

• You have a platform engineering team that wants to own agent infrastructure
• You are standardized on Google Cloud and want agents close to BigQuery and Pub/Sub
• You are building customer-facing or product agents, not back-office operations
• You want to design your own guardrails, approvals, and audit model

Choose Runtime if

• Payment ops, risk, compliance, or finance teams need agents working real queues soon
• Approvals before money moves and an examiner-ready audit trail are requirements
• You run on more than one cloud, or want to bring your own sandbox and models
• You want a forward-deployed engineer who has built payment infrastructure, not a toolkit

You can also run both: engineers keep building product agents on Google Cloud, and operations runs on Runtime in the same GCP account. Rain replaced $250k in vendor spend with Runtime.