Which Major Platforms Have Shipped MCP Servers in 2026?

Eighteen months ago, MCP was a protocol with a handful of community-built servers and a lot of promise. Today, every major cloud provider and most of the top SaaS platforms run official MCP servers in production. If you are building AI agents, this changes what is possible out of the box.

Here is a breakdown of who shipped what, when they did it, and what it means for the developer ecosystem.

The Hyperscalers

AWS

AWS made its MCP server generally available on May 6, 2026. Agents can call any AWS API through a single MCP tool, run sandboxed Python, and discover documentation without needing credentials baked into prompts. IAM-based guardrails, CloudWatch metrics, and CloudTrail logging give teams the visibility they need for production use.

Google Cloud

Google went further than anyone expected at Cloud Next 2026. They shipped Workspace MCP (Gmail, Docs, Sheets, Calendar), managed database MCP servers for Spanner and BigQuery, BYO-MCP support in Vertex AI, and an Agents CLI for local development. MCP is now a first-class primitive in the Google Cloud stack.

Microsoft

Microsoft integrated MCP into Copilot Studio in July 2025, making it one of the earliest major platform adopters after Anthropic. This lets enterprise teams connect Copilot agents to any MCP-compatible server without writing custom connector code.

Developer Platforms

GitHub

GitHub runs an official remote MCP server that gives agents access to repositories, pull requests, issues, and Actions workflows. For AI coding assistants, this is the most natural integration point in the development loop.

Figma

Figma’s hosted MCP server lives at mcp.figma.com. Design agents can read files, inspect components, and extract design tokens. For teams building design-to-code pipelines, this removes the need for screenshot-based workarounds.

Linear

Linear ships an MCP server that exposes issues, projects, and cycles. Agents can create, update, and query work items directly. If your team already uses Linear for project tracking, an agent can now participate in the same workflow without API key plumbing.

Commerce and Payments

Stripe

Stripe runs an official remote MCP server at mcp.stripe.com. Agents can create payment links, look up charges, manage subscriptions, and query balance transactions. For teams building agent-powered checkout or billing flows, this is a direct path to production.

Shopify

Shopify went wide with four official MCP servers: Storefront, Customer Account, Checkout (still in preview), and Dev. This covers the full merchant stack, from product catalog queries to order management to theme development.

Business Software

Salesforce

Salesforce ships an MCP server that exposes CRM objects, reports, and flows. Agents can query contacts, update opportunities, and trigger automations. Given Salesforce’s installed base, this single integration opens up a massive surface area for enterprise agent builders.

HubSpot

HubSpot followed with its own MCP server covering contacts, deals, tickets, and marketing workflows. The pattern is the same: instead of building against a REST API, agents discover available tools through the MCP handshake and call them with structured input.

Notion

Notion’s MCP server gives agents read and write access to pages, databases, and blocks. For teams using Notion as a knowledge base, this turns it into a live data source for agents rather than a static reference.

Slack

Slack’s MCP server exposes channels, messages, and user presence. Agents can post messages, search history, and respond to threads. Combined with other MCP servers, this enables multi-step workflows where an agent queries data from one system and reports findings in Slack.

What the Numbers Say

The scale of adoption is hard to overstate. As of Q2 2026:

  • 9,400+ servers are published across the four major registries
  • 97 million monthly downloads of the Python and TypeScript SDKs combined
  • 78% of enterprise AI teams report at least one MCP-backed agent in production
  • 28% of Fortune 500 companies have deployed MCP servers for production AI workflows

The shift from local stdio servers to remote HTTP deployments has been just as significant. Remote server deployments grew nearly 4x since May 2025. For enterprise teams, remote servers solve the problems that blocked adoption: shared access, centralized auth, session persistence, and independent scaling.

What This Means for Builders

Two years ago, connecting an AI agent to Stripe meant writing a custom integration, handling auth, parsing responses, and maintaining it as the API evolved. Today, you point your agent at mcp.stripe.com and it discovers what it can do.

This is not just convenience. It changes the economics of agent development. A solo developer can now wire an agent into AWS, Stripe, GitHub, Slack, and Notion in an afternoon. The integration layer that used to take weeks is now a configuration step.

If you are building agents and want to find the right MCP servers for your stack, the AgentNDX directory indexes thousands of servers with filtering by category, protocol version, and deployment type.

FAQ

Do all these platforms offer remote MCP servers?

Most of the major platforms listed here run hosted remote servers using the Streamable HTTP transport. Some also support local stdio deployment for development use. The trend is strongly toward remote-first.

Is there a single registry where I can find all official MCP servers?

Anthropic launched an official MCP registry at registry.modelcontextprotocol.io. It does not cover every server, but it is the closest thing to a canonical source. AgentNDX indexes servers from multiple registries and community sources.

Do I need OAuth to connect to these servers?

The MCP spec added OAuth 2.1 authorization in April 2026. Most enterprise MCP servers require OAuth-based auth. Some developer-focused servers still support API key auth for simpler setups.

Which MCP server should I start with?

That depends on your use case. For a coding agent, start with GitHub. For a business automation agent, Slack and Notion are natural first choices. For anything touching payments, Stripe’s server is production-ready today.