Google just made its Gemini Business Edition a lot more useful for companies that don’t want their AI assistant living in a walled garden. Team administrators can now configure custom Model Context Protocol (MCP) server connections, letting the platform talk directly to private company data, internal tools, and third-party systems that support the protocol.
Think of MCP as a universal adapter for AI. Instead of building bespoke integrations every time you want an AI model to access a new tool or dataset, MCP provides a standardized way for AI applications and external systems to shake hands. Anthropic originally launched the protocol in late 2024, and it has since gained traction across the industry as a de facto connectivity standard.
What Google actually shipped
Google first dipped its toes into MCP waters back in December 2025, when it announced support for managed remote MCP servers across its services. That was helpful, but it only covered pre-configured, Google-managed connections. If your company had a legacy inventory system or a proprietary analytics tool that wasn’t on Google’s menu, you were out of luck.
The new custom MCP server capability fills that gap. Administrators can now point Gemini Business Edition at their own MCP-compliant servers, connecting the AI platform to whatever internal infrastructure a company already runs.
There are some guardrails, though. Google’s implementation only supports StreamableHTTP transport over HTTPS. Server-Sent Events, the other common transport mechanism in the MCP ecosystem, are explicitly not allowed. Every connection must use publicly trusted TLS certificates, which means no self-signed shortcuts that might give a security team heartburn.
Why MCP matters for enterprise AI
For enterprises, the practical upside is significant. A company running Gemini Business Edition can now have its AI assistant pull data from internal systems during conversations, trigger actions in proprietary tools, or access information stored in databases that would never be exposed to a public API. All of this happens within the security framework that enterprise IT teams demand, with OAuth 2.0 authentication and TLS encryption baked into the requirements.
The competitive landscape
Google’s documentation for the feature is slated to remain accessible through September 2026, which signals this isn’t an experimental beta destined for the deprecation graveyard. The move also positions Google to capture a growing segment of the enterprise software market: companies that have already invested in MCP-compliant infrastructure for other AI tools and want to extend those investments to Gemini.