Which Major Platforms Have Shipped MCP Servers in 2026?

A year ago, MCP was a developer-community project. The servers were built by independent contributors, open-source teams, and small startups wrapping third-party APIs. The major platforms watched from the sidelines.

That changed fast. In 2026, the biggest names in cloud, developer tooling, and SaaS started shipping their own MCP servers. Not wrappers built by community members. Official, maintained, first-party integrations. Here is who shipped what, and why it matters.

The Hyperscalers

Google Cloud

Google made the most aggressive move at Cloud Next 2026, announcing MCP as a first-class cloud primitive. The rollout included a Workspace MCP Server (Drive, Gmail, Calendar, Chat), managed remote MCP servers for Cloud SQL, Spanner, and AlloyDB, and a BYO-MCP hosting layer for custom servers. Google also shipped Google Maps MCP and Google Drive MCP as standalone servers in the directory.

The bet is clear: openness drives Workspace adoption. Any agent that speaks MCP can now plug into Google’s productivity suite, not just Gemini.

AWS

Amazon shipped the AWS API MCP server, then followed with AWS Documentation MCP and AWS Secrets Manager MCP. When the AWS MCP server went generally available, enterprise teams finally had a reason to care about the protocol.

Different philosophy than Google. No single managed platform — just modular servers, one per service. Classic AWS: each team owns its own integration.

Microsoft

Microsoft entered with Microsoft Teams MCP, connecting agents to Teams channels, chats, and meetings. Combined with GitHub’s MCP server (which Microsoft owns), the company now has MCP coverage across both developer and enterprise communication workflows.

Developer Tooling

GitHub

GitHub MCP was one of the earliest platform-backed servers — repos, issues, PRs, code search, and actions, all through MCP tools. At this point it is table stakes. Nearly every coding agent that supports MCP connects to GitHub first.

Docker

Docker MCP lets agents manage containers, images, and compose stacks. If an agent is building and deploying software, it needs container access. Docker shipping an official server killed off the fragile community wrappers overnight.

Vercel

Vercel MCP gives agents access to deployments, projects, and environment configuration. It closes the gap between generating frontend code and actually shipping it.

Sentry

Sentry MCP connects agents to error tracking, performance monitoring, and release management. The practical upside is obvious: an agent debugging an issue pulls real error data from production without leaving context.

Databases and Data Infrastructure

Supabase

Supabase MCP bundles Postgres, auth, storage, and edge functions into a single server. It quickly became one of the most-installed servers in the directory — no surprise, since the Supabase developer crowd and early MCP adopters are basically the same people. We wrote a full spotlight on it.

MongoDB

MongoDB MCP provides document database operations through the protocol. Queries, aggregations, index management, and collection operations are all available as MCP tools.

Redis

Redis MCP handles key-value operations, pub/sub, and caching through MCP. Agents that need fast state or inter-agent messaging have a clean path now.

Snowflake

Snowflake MCP opens up data warehouse queries and transformations. Enterprise data teams use it to let agents run analytics without handing out direct SQL access to production.

SaaS and Business Tools

Stripe

The Stripe Agent Toolkit stood out. We wrote a full spotlight on it. Stripe didn’t just expose payment APIs through MCP — they built an opinionated toolkit around how agents actually use payments: checkout sessions, subscription management, refunds, transaction reads. It’s shaped by real agent behavior, not just API surface area.

Salesforce

Salesforce MCP opens CRM data to agents — leads, contacts, opportunities, custom objects. If you’re building a sales or support agent, Salesforce is probably the single biggest context source it needs.

Shopify

Shopify MCP connects agents to storefronts, products, orders, and inventory. E-commerce agent automation is still early, but at least the integration path is no longer the hard part.

Notion

Notion MCP — pages, databases, blocks. One of the most common integrations we see. If an agent manages tasks or documentation, Notion is usually in the stack.

Linear

Linear MCP exposes issues, projects, cycles, and teams. Pair it with GitHub MCP and an engineering agent has the full loop — code and project tracking in one context.

Slack

Slack MCP connects agents to channels, messages, and threads. Multi-agent systems that need to report results or loop in humans tend to land here.

Figma

Figma MCP — design files, components, styles. Narrow use case, but when an agent generating frontend code can reference actual design specs instead of guessing, the output quality jumps.

Monitoring and Observability

Datadog

Datadog MCP provides access to metrics, logs, traces, and monitors. An agent handling an incident at 2am can pull real observability data instead of waiting for a human to paste it into a prompt.

Grafana

Grafana MCP connects agents to dashboards, data sources, and alerting. If your observability stack runs on Grafana, this is how agents plug into it.

Infrastructure and Security

Cloudflare

Cloudflare MCP exposes Workers, Pages, DNS, and security configs. One less dashboard an agent (or its operator) has to touch.

Twilio

Twilio MCP — SMS, voice, and communication APIs. When your agent needs to actually reach a customer (not just update a database), Twilio is the standard path.

Search and Web

Brave Search MCP provides web search without Google’s API pricing. It has quietly become one of the default search integrations in the agent space.

Exa

Exa MCP takes a different angle — search built from the ground up for agents, with semantic queries and structured results. We covered it in our Exa spotlight.

Firecrawl

Firecrawl MCP handles web scraping, crawling, and structured data extraction. We use it ourselves. See our Firecrawl spotlight.

What this means

The obvious thing: maintenance. Community servers go stale when their maintainers lose interest. Platform-backed servers have teams behind them. When the underlying API changes, the MCP server gets updated. That alone matters.

But the bigger shift is trust. A company that would never deploy a community MCP server from some random GitHub repo will happily deploy one from their existing vendor. First-party auth helps too — Google’s managed servers inherit IAM, Stripe’s toolkit uses existing API keys. No extra credential management, no extra procurement review.

We’re tracking over 900 curated servers in the AgentNDX directory now. The share backed by major platforms keeps growing every month. A year from now, an MCP server from a major platform will be expected, not noteworthy.

FAQ

Q: Do I need to use a platform’s official MCP server, or can I use a community-built alternative?

A: Both work. Official servers tend to have better maintenance and auth integration. Community servers sometimes cover more features or niche use cases. Check the AgentNDX directory to compare what is available for any given platform.

Q: Which platform was first to ship an official MCP server?

A: GitHub was among the earliest, followed closely by Stripe and Cloudflare. Google made the biggest single announcement at Cloud Next 2026 with multiple servers and managed MCP infrastructure.

Q: Are these servers free to use?

A: The MCP servers themselves are generally open source or free to run. You still need credentials (API keys, OAuth tokens) for the underlying platform, and the platform’s normal pricing applies to API calls made through the server.