Google Just Made AI Agents Plug-and-Play

On December 10, Google launched fully managed MCP servers for its cloud services. If you build AI agents, this is a big deal. If you don’t, here’s why it should still be on your radar.

What’s MCP?

Model Context Protocol is an open standard Anthropic released in late 2024. Think of it as a universal connector between AI models and external tools — like USB-C, but for AI. Instead of building custom integrations for every service, developers use MCP to let their agents “discover” and use tools automatically.

The problem before MCP: every AI integration was bespoke. Want your agent to query a database? Write custom code. Want it to pull from Google Maps? Different custom code. Want it to manage Kubernetes? Yet another integration. Every new tool meant weeks of development and maintenance.

MCP standardizes that entire layer. One protocol, any tool, any model.

How Fast MCP Took Over

The adoption numbers tell the story. According to MCP ecosystem tracking data, MCP server downloads grew from roughly 100,000 in November 2024 to over 8 million by April 2025. In under six months, the ecosystem went from experimental to industry standard.

As of mid-2025, registries like PulseMCP and Glama list over 5,800 MCP servers and 300 MCP clients, spanning data sources, APIs, and enterprise tools. Major companies have built MCP servers: Notion for note management, Stripe for payment workflows, GitHub for engineering automation, Hugging Face for model management, and Postman for API testing.

OpenAI adopted MCP across its Agents SDK, Responses API, and ChatGPT desktop app in March 2025. Google DeepMind confirmed MCP support for Gemini models in April 2025. When the two biggest players in AI both adopt the same open protocol within months of each other, that’s not a trend — it’s a standard.

What Google Just Did

According to TechCrunch, Google is now offering fully managed, remote MCP servers for Maps, BigQuery, Compute Engine, and Kubernetes Engine. No infrastructure to maintain. Just paste a URL and your agent can talk to Google’s services.

The servers are offered at no extra cost to enterprise customers who already pay for Google Cloud services. That pricing decision matters — it signals Google sees MCP as infrastructure, not a product to monetize separately.

“We are making Google agent-ready by design,” said Steren Giannini, Google Cloud’s product management director, in an interview with TechCrunch.

Critically, these MCP servers work with any AI model that supports the protocol. Claude, Gemini, GPT-5, and open-weight models like Devstral can all access Google Cloud services through the same interfaces. Google isn’t locking this to Gemini — they’re building infrastructure for the entire ecosystem.

Why This Matters for Small Businesses

If you’re not building AI agents yourself, this still affects you. Here’s why.

Every SaaS tool you use is about to get smarter. When the protocol connecting AI models to business tools becomes standardized, the software you already pay for will start offering AI features that actually work across your stack — not just within one app’s walled garden.

The practical example: imagine an AI assistant that can check your Google Maps listings, pull sales data from BigQuery, and spin up a test server on Compute Engine — all from a single conversation. Before MCP, that required three separate integrations. Now it requires one protocol.

For consultancies and agencies, this is a direct cost savings. Building AI agent integrations used to mean weeks of custom development per service. MCP reduces that to configuration, not code.

The Security Layer

Enterprise adoption of AI agents has been slow for one reason: security. Giving an AI agent access to production infrastructure is terrifying without proper guardrails.

Google addressed this head-on. Model Armor firewall is enabled by default on all managed MCP servers, providing input validation, action classification, and behavioral analysis for AI agent operations. Administrators manage access through Google Cloud IAM with full audit logging.

The official announcement also highlights Apigee integration — Google’s API management product can essentially “translate” any standard API into an MCP server. That means companies can expose their own internal APIs as tools that AI agents can discover and use, with existing security and governance controls layered on top.

Anthropic’s David Soria Parra, co-creator of MCP, said Google’s support “brings us closer to agentic AI that works seamlessly across the tools and services people already use,” per the Google Cloud blog.

What’s Coming Next

Google says they’ll expand MCP support to Cloud Storage, Cloud SQL, Spanner, Looker, Pub/Sub, and more in the coming months. Database support is expanding to include PostgreSQL with AlloyDB, Spanner and Cloud SQL for relational workloads, plus Firestore and Bigtable for high-performance NoSQL.

Beyond Google, the broader MCP ecosystem continues to mature rapidly. The protocol’s 2026 roadmap prioritizes enterprise readiness — including audit trails, SSO-integrated auth, and gateway behavior — alongside stable transport layers and governance frameworks that will enable wider production adoption.

The Scramble Take

This validates MCP as the industry standard for agent connectivity — which matters because standards create ecosystems. When Google, Anthropic, OpenAI, and Microsoft are all aligned on the same protocol, it becomes much easier to build agents that work everywhere.

The practical impact: expect a lot more AI agents that can actually do things — query databases, manage infrastructure, pull location data — without developers reinventing the wheel for every integration. The agent era just got its infrastructure layer.

For small businesses and solopreneurs, the takeaway is simpler: the tools connecting AI to your business are getting cheaper, faster, and more standardized. The window to learn this stuff before your competitors do is closing.