Remote MCP servers hit production scale
Ten leading AI companies — Anthropic, Asana, Atlassian, Block, Intercom, Linear, PayPal, Sentry, Stripe, and Webflow — have deployed remote Model Context Protocol (MCP) servers on Cloudflare. The new servers let Claude users manage projects, generate invoices, query databases, and deploy full-stack applications directly from the chat interface.
Since Anthropic introduced MCP in November, the protocol has emerged as the standard way for AI agents to interact with external tools. Until now, however, users had to install MCP servers locally. With Anthropic's new Integrations announcement, remote MCP servers can be accessed by URL, no local installation required.
Cloudflare's tooling abstracts away protocol complexity, authentication, and transport handling, so engineering teams can focus on building MCP tools for their applications. Deploying a pre-built MCP server — with support for the latest MCP standards — takes a single click. A "Deploy to Cloudflare" button is available for a quickstart template:
The new servers are being showcased at Demo Day on May 1 at 10:00 AM PST. Cloudflare is also making its own remote MCP servers available to customers, along with lessons learned from building them.
The MCP transition is underway
Just as every business eventually needed a website as internet access spread, every service is now moving toward AI-native access. Developers increasingly start coding sessions with tools like Claude — and it follows that the same assistant should help deploy the code it helped write. Non-developers, similarly, should be able to move from asking Claude for a recipe to having it order the groceries.
Remote MCP makes these workflows possible. And as with the early web, first movers building MCP-forward services stand to gain an advantage: the faster users reach the value of a product, the more likely they are to succeed, upgrade, and stay.
What MCP unlocks for businesses
Adopting MCP gives service providers several concrete benefits:
- Lower barrier to entry: Users don't need to learn a dashboard or read documentation. They describe what they want, and the agent figures out how to accomplish it.
- Personalized experiences: MCP tracks a user's requests and interactions over time, allowing future tool calls to adapt to individual usage patterns and preferences.
- Natural upgrade paths: Instead of feature comparison tables, the AI agent can explain how a higher-tier plan helps a specific user accomplish their particular goals.
- New feature and integration channels: Exposing tools via MCP lets users connect a product to the rest of their stack. Agents can combine tools across providers, enabling integrations without building every third-party connection in-house.
Why Cloudflare for MCP servers
Since MCP's launch, Cloudflare has focused on simplifying remote MCP server development. The platform abstracts protocol complexity, handles authentication and transport, and supports fully stateful MCP servers by default. Stateful servers can "sleep" when idle to minimize costs.
Recent updates include:
- Streamable HTTP transport support: MCP servers can now use the latest Streamable HTTP transport alongside SSE. The AI Playground and
mcp-remoteproxy have been updated to provide a remote MCP client for testing the new transport. - Python support: MCP servers can now be built on Cloudflare using Python, in addition to JavaScript and TypeScript.
- Expanded developer resources: New quickstart templates, updated documentation covering the new transport method, and published best practices for building MCP servers.
Rather than describing what these capabilities look like in practice, the company partners' deployments demonstrate the production use cases. The new MCP servers are available to Claude users today through Anthropic's Integrations feature.
Production MCP servers from major vendors
Several companies have already moved from testing MCP to running first-party remote MCP servers in production. These are live integrations on Cloudflare’s infrastructure that customers can connect to today.
Asana
Asana’s MCP server gives users natural-language access to the Asana Work Graph from any MCP client. Users can get project updates, search tasks, manage projects, post comments, and change deadlines. The broader goal is agent-driven interoperability: turning meeting notes from another app into an assigned set of Asana tasks, or pulling Asana tasks into an MCP-enabled IDE. Asana credits Cloudflare’s foundational tooling with letting the team ship its first-party server quickly and handle authentication and scaling while they focused on the user experience.
Atlassian
The Atlassian Remote MCP Server, now in beta, lets Jira and Confluence Cloud customers work with their data from Anthropic’s Claude app. It runs on Cloudflare infrastructure and keeps data within permissioned boundaries. Users can summarize work items or pages, create issues or pages directly from Claude, perform bulk actions, and enrich Jira items with context from other sources Claude can access. Atlassian says Cloudflare provided everything from OAuth to out-of-the-box remote MCP support, allowing a fully operational setup to be built, secured, and scaled quickly.
Intercom
Intercom’s Fin AI agent now resolves over half of support conversations autonomously for customers like Anthropic. The Intercom MCP server unlocks conversation and user data that was previously siloed in the support platform, making it available to anyone in the organization using AI tools. Engineers can use conversation history in tools like Cursor or Claude Code to diagnose and fix issues faster. Intercom’s engineering leadership reports going from concept to production in under a day using Cloudflare’s toolkit, and plans to encourage customers to build their own MCP servers on the same platform.
Linear
The Linear MCP server brings issue tracking and product development context directly into AI assistants when it’s needed — refining a product spec in Claude, collaborating on a bug fix in Cursor, or creating issues from an email. Linear chose Cloudflare for its frameworks in a fast-moving space plus flexible edge compute. The goal is to eliminate context switching so developers and product teams can access their work where they already are.
PayPal
PayPal’s remote MCP server lets developers delegate commerce workflows to AI agents using natural language. The server exposes PayPal’s commerce capabilities — inventory management, payment processing, shipping tracking, and refunds — so agents can autonomously execute and optimize commerce operations. PayPal describes MCP as a new paradigm for software development and notes that developers can begin integrating today.
Sentry
Sentry’s MCP server lets developers query error and issue context from an IDE or an assistant like Claude, across projects or for individual files. It can also create projects, capture setup information, and query project and organization details. Sentry plans to extend the MCP with root cause analysis from its Seer agent and simplified instrumentation for traces and exception handling. Building on Cloudflare’s Remote MCP let Sentry sidestep local scaling and authentication complications; the company leveraged its own OAuth configuration and used Durable Objects to maintain state outside a local environment.
Block
Square’s API suite covers payments, orders, inventory, and customer management. The dedicated Square MCP server makes all of those API resources available to AI agents like Claude and Block’s codename goose. Sellers can build customized use cases across Square’s full capabilities at a lower technical barrier than direct API integration.
Stripe
Stripe sees MCP emerging as a new AI interface — potentially the default or even the only way people, businesses, and code discover and interact with services. With Stripe’s agent toolkit and Cloudflare’s Agent SDK, developers can monetize their MCPs with just a few lines of code, according to Stripe’s product lead.
Webflow
Webflow’s MCP server supports CMS management, SEO auditing and improvement, content localization, and site publishing. Users can manage their site directly from an AI agent. Webflow treats MCP not as a different surface over its APIs but as a way to expose discrete actions — publish a site, create a CMS collection, update content — in a discoverable, secure, and contextual way. The company cited Cloudflare’s mature remote MCP support and its alignment on authentication and durability as reasons for building on the platform.
Getting started with remote MCP
Developers who want to build a remote MCP server can start with Cloudflare’s documentation, watch a tutorial video, or use the starter template to get a server deployed to production. A deployed remote MCP server works from Claude, Cloudflare’s AI playground, or any remote MCP client.
A new YouTube video walks through building MCP servers using two of Cloudflare’s MCP templates.
Questions and feedback can be sent to [email protected].
