This spotlight is part of our MCP company spotlights, within The MCP Guide. Building an MCP company? See SEO and GEO for MCP companies.
Google supports the Model Context Protocol (MCP) as a client, a server publisher and a standards backer. Gemini products can connect to MCP servers, Google Cloud runs managed MCP endpoints for its services, and Google co-maintains the official Go SDK. Google also created A2A, the agent-to-agent protocol that now sits beside MCP in the same foundation.
Checked against: Google Cloud MCP servers overview on 28 October 2026. eSEOspace did not test Google's products.
Key Takeaways
- Google is a Platinum member of the Agentic AI Foundation (AAIF), which hosts both MCP and A2A.
- The consumer Gemini web app now lets eligible US users add "custom apps" by MCP server URL, per Google's help center.
- Gemini CLI and Gemini Enterprise both document MCP support, with different setup paths.
- Google Cloud says more than 50 Google-managed MCP servers are generally available or in preview.
- MCP Toolbox for Databases and the Go SDK are open source under Apache 2.0.
At a glance
| Item | Detail |
|---|---|
| Company | |
| What it offers for MCP | MCP clients (Gemini app, Gemini CLI, Gemini Enterprise), managed MCP servers, MCP Toolbox, Go SDK |
| Who it's for | Google Cloud and Workspace teams, developers, eligible consumer Gemini users |
| Open source? | Partly: Gemini CLI, MCP Toolbox and Go SDK are open source; managed servers are hosted by Google |
| Pricing | No separate MCP price listed; depends on the underlying product |
What Google offers for MCP
Governance and protocols
The AAIF About page lists Google among eight Platinum members. April Kyle Nassi of Google is Treasurer of the Governing Board. Our guide on who owns MCP explains how that board works.
According to the AAIF announcement dated 17 August 2026, "Google launched A2A in April 2025 and donated it to the Linux Foundation." A2A then joined AAIF as a hosted project. Our guide to MCP vs A2A vs function calling explains how they differ.
The official Go SDK says it is "Maintained in collaboration with Google." New contributions use Apache 2.0, and older code stays under MIT.
MCP clients
- Gemini app. Google's help article, Connect & manage custom apps for Gemini Apps, describes adding third-party MCP servers by URL. You must connect them in the web app, though they then work on mobile too.
- Gemini CLI. The Gemini CLI MCP docs cover stdio, SSE and Streamable HTTP servers, plus tools, resources, prompts and OAuth. The Gemini CLI repository says free individual access and Google AI Pro and Ultra plans moved to Antigravity CLI on 18 June 2026.
- Gemini Enterprise. Admins can connect a custom MCP server as a data store. It supports only Streamable HTTP, and "all actions are turned off" by default.
MCP servers and tools
- Google-managed MCP servers. A Google Cloud blog post from 28 April 2026 says "more than 50" are generally available or in preview. Examples include BigQuery, Cloud Run, GKE, Spanner and Workspace apps. Google says its remote servers "support MCP version 2026-07-28." See our Google Cloud MCP profile.
- MCP Toolbox for Databases. The MCP Toolbox repository calls it "an open source Model Context Protocol (MCP) server" for databases. It covers Google Cloud databases plus PostgreSQL, MySQL, MongoDB, Snowflake and others.
Who it's for
Google's MCP work fits teams already on Google Cloud or Google Workspace. Developers can use Gemini CLI, MCP Toolbox and the managed endpoints. Business teams on Gemini Enterprise can connect internal servers through an admin. Individual US users with personal accounts can try custom apps in the Gemini web app.
Pricing
No separate MCP price listed. MCP itself is free and open source under its license. The Google Cloud overview and launch blog post we checked make no pricing statement, so cost depends on the underlying product. MCP Toolbox, Gemini CLI and the Go SDK are open source under Apache 2.0 (the Go SDK keeps older code under MIT). The Gemini help article for custom apps lists eligibility rules, not a plan requirement.
Security and trust
Google says Cloud remote servers use IAM to control access. It adds that Model Armor "helps secure your agentic AI applications by sanitizing MCP requests and responses." The launch post also mentions IAM Deny policies, OpenTelemetry tracing and Cloud Audit Logs.
For the Gemini app, Google warns: "Google does not control, monitor, or secure third-party MCP servers." It says Gemini "requires manual confirmation for any write actions, but Gemini can make mistakes." Our guide to MCP security risks covers what to check before you connect a server.
Alternatives
- Microsoft and MCP for Microsoft 365, Azure and GitHub teams.
- OpenAI and MCP for ChatGPT and the Responses API.
- Which AI apps support MCP for a wider comparison.
Next step
Pick the Google surface your team already uses, then connect one low-risk, read-only server first. Browse more company profiles in the MCP Guide company directory, or talk to eSEOspace about MCP development for a server your Gemini users can reach.
Frequently asked questions
Can I add my own MCP server to the Gemini app?
Yes, if you are eligible. Google's help center says you must be 18 or over, in the US, signed in with a personal Google Account and have Keep Activity on. Work and school accounts are not supported, and custom apps are English only.
Does Gemini CLI still support MCP?
Yes. Google's Gemini CLI docs describe stdio, SSE and Streamable HTTP servers. Free individual users and Google AI Pro and Ultra plans moved to Antigravity CLI on 18 June 2026, per the Gemini CLI repository.
What is A2A, and is it part of MCP?
A2A (Agent2Agent) is a separate protocol for agents talking to other agents. Google launched it in April 2025. It is now an AAIF hosted project, alongside MCP, but it is not part of the MCP spec.
Are Google Cloud MCP servers free?
Google's MCP overview and launch post do not state a price, so cost depends on the underlying product, such as BigQuery. Check each product's pricing page before you roll out.
