#### Videos
Use the HTML video element for self-hosted video content:
Embed YouTube videos using iframe elements:
#### Tooltips
Example of tooltip usage:
***
## 2) Add the Cloud MCP Connector to Claude (Remote HTTP MCP)
1. In Claude, open **Settings → Connectors**.
2. Choose **Add Custom Connector**.
3. Paste the Cloud MCP server URL (the OAuth-protected endpoint that Cloud MCP provides).
4. Approve the **Cloud MCP OAuth** screen when prompted.
5. After a successful handshake, you should see the tools listed in the chat window; expand it to confirm.
> If tools don’t appear: re-open the connection panel, confirm URL in Cloud MCP, and retry the OAuth approval.
***
## 4) Use your Cloud MCP Claude Connector Tools!
***
## 6) Troubleshooting
* **MCP tools don’t appear in Claude:** Reconnect the Remote (HTTP) server in Settings; confirm the Cloud MCP server is **Healthy** and the OAuth grant is still valid.
* **Connector not available in chat:** Ensure it’s toggled on in **Settings → Connectors** and that you’ve authorized it.
* **Ambiguous tool routing:** In your prompt, explicitly say *which* tool to use for *which* sub-task. Avoid enabling unnecessary connectors/tools for a session.
***
## 7) Privacy & access notes
* Cloud MCP keeps your MCP server behind **OAuth**; Claude receives only tool responses.
* Disable any connector you don’t need for a session.
***
## 8) Where to go next
* Add more MCP servers to Cloud MCP (e.g., research, code intelligence, browsers).
* Create **small, purpose-built** prompts that map cleanly to a tool or connector.
* Save a “starter chat” template in your team wiki with the exact toggles and prompt you use for repeatable tasks.
### Need Help?
If you run into any issues:
* Reach out to [me](mailto:vikash@cloudmcp.run)
* Check our [Blog Posts](https://cloudmcp.run/blog)
* Join our [Discord community](https://discord.gg/cloudmcp)
***
*Ready to deploy your Claude Connectors? [Sign Up for a plan! →](https://cloudmcp.run/pricing?utm_source=blog\&utm_medium=claude-connectors-guide)*
# How to Deploy a Remote MCP Server (on cloudmcp.run)
Source: https://docs.cloudmcp.run/blog/cloud-mcp-deployment-guide
Learn how to deploy Model Context Protocol (MCP) servers on cloudmcp.run in just a few clicks. This step-by-step guide walks you through finding, saving, and deploying MCP servers.
# How to Deploy an MCP Server on CloudMCP.run
Deploying an MCP (Model Context Protocol) server on CloudMCP.run is designed to be quick and straightforward. In this guide, we'll walk you through the entire process from finding a server to having it deployed and running in production.
## Prerequisites
Before you begin, make sure you have:
* A cloudmcp.run account (sign up with GitHub at [cloudmcp.run](https://cloudmcp.run))
* Basic understanding of what MCP servers do
* Any required API keys or credentials for the server you want to deploy
## Step 1: Find Your MCP Server
Start by navigating to the "Find Servers" page from your dashboard. Here you can browse our registry of available MCP servers or import your own from a GitHub repository.
You can:
* **Search the registry**: Type keywords to find servers that match your needs (e.g., "desktop-commander", "filesystem", "database")
* **Import your own**: Click the "Import Server" button to link your own GitHub repository
* **Bookmark/Save Servers**: Click the bookmark icon on any server card to save it to your personal library for easy access later
For this guide, we'll deploy the popular `@wonderwhy-er/desktop-commander` server, which gives Claude terminal control and file system access.
## Step 2: Browse and Save Servers
Once you've found servers you're interested in, they'll appear in your "Saved Servers" section. This is your personal library of MCP servers ready to deploy.
From this view, you can:
* See your deployed servers
* See all your saved servers with their descriptions
* Check the package name and version
* View when each server was saved
* Click "Deploy Server" to begin deployment
Each server card shows:
* **Package name**: The npm package or GitHub repository
* **Version**: Current version or branch
* **Description**: What the server does
* **Save date**: When you added it to your library
## Step 3: Configure Your Deployment
Click "Deploy Server" on your chosen server to open the deployment configuration modal.
### Configuration Options:
#### Deployment Name
Give your deployment a unique, descriptive name. This helps you identify it later, especially if you deploy the same server multiple times with different configurations.
#### Command
The command to run your server. Common options:
* `npx` - For npm packages (most common)
* `node` - For custom scripts
* `python` - For Python-based servers
#### Arguments
The specific package or script to run. For npm packages, this is typically:
* `-y @package-name` (the `-y` flag skips npm prompts)
#### Environment Variables
Add any required configuration:
* API keys
* Database connection strings
* Custom settings
* Authentication tokens
> 🔒 **Security Tip**: Environment variables are encrypted and never exposed in logs or the UI after creation.
### Example Configuration
For the desktop-commander server:
* **Deployment Name**: `desktop-commander`
* **Command**: `npx`
* **Arguments**: `-y @wonderwhy-er/desktop-commander`
* **Environment Variables**: (Add any required API keys or settings)
Click "Deploy" when you're ready to launch your server.
## Step 4: Monitor Your Deployment
After clicking deploy, you'll be redirected to your deployments page where you can monitor the status of your server.
### Deployment Details:
Your deployed server card shows:
* **Status indicator**: Green dot for active, yellow for deploying, red for stopped
* **Deployment date**: When the server was deployed
* **Description**: What the server does
* **Package info**: The underlying package being run
* **Server URL**: Your unique endpoint (e.g., `https://2976130-e28603...`)
### Available Actions:
* **Copy URL**: Click the copy icon next to the server URL to copy it to your clipboard
* **Install in VSCode**: 1-click install directly into VSCode with OAuth authentication
* **Stop**: Shut down and delete the server
## Step 5: Connect Your MCP Client
Now that your server is deployed, you can connect to it from any MCP-compatible client.
### For Configuration-file Based Clients:
1. Copy your server URL from the deployment card
2. In your MCP client configuration, add:
```json theme={null}
{
"mcpServers": {
"desktop-commander": {
"url": "https://your-server-url.cloudmcp.run"
}
}
}
```
3. Restart your client to connect to the server
### Testing Your Connection
Once connected, you can test your server:
* For desktop-commander: Try asking Claude to list files or run terminal commands
* For database servers: Query your connected database
* For API servers: Make requests to your configured endpoints
## Managing Your Deployments
Click on your deployment card to view:
* 1-click VSCode MCP installation.
* Performance metrics\* (soon!)
* Stop action
### Scaling and Performance
CloudMCP.run automatically handles:
* **Auto-scaling**: Servers scale based on demand
* **Load balancing**: Requests are distributed efficiently
* **Failover**: Automatic recovery from crashes
* **Caching**: Improved response times for repeated requests
## Best Practices
1. **Use descriptive names**: Make it easy to identify deployments
2. **Stop unused servers**: Save resources by stopping inactive deployments
## Troubleshooting
### Server Won't Start
* Check your environment variables are correctly formatted
* Verify the package name and version
* Review deployment logs for error messages
### Connection Issues
* Ensure your server URL is correctly copied
* Check that your client supports the MCP version
* Verify network connectivity
### Performance Problems
* Check server logs for errors
* Consider upgrading your plan for more resources
* Contact support if issues persist
## Advanced Features
### Multiple Deployments
You can deploy the same server multiple times with different configurations:
* Different environment variables
* Separate instances for development/production
* Isolated deployments for different projects
### Custom Servers
Import your own MCP servers from GitHub:
1. Click "Import Server" on the Find Servers page
2. Enter your GitHub repository URL
3. Configure and deploy like any other server
## What's Next?
Now that you've successfully deployed your first MCP server:
* 🚀 **Deploy more servers**: Explore our registry for other useful integrations
* 📚 **Read the docs**: Deep dive into advanced configuration options
* 💬 **Join our Discord**: Connect with other developers using CloudMCP
* 🛠️ **Build your own**: Create custom MCP servers for your specific needs
## Need Help?
If you run into any issues:
* Check our [documentation](https://docs.cloudmcp.run)
* Join our [Discord community](https://discord.gg/cloudmcp)
* Email support at [support@cloudmcp.run](mailto:support@cloudmcp.run)
***
*Ready to deploy your next MCP server? [Head to your dashboard →](https://cloudmcp.run/dashboard)*
# Introducing Cloud MCP Router: Progressive Tool Discovery for Real‑World AI Agents
Source: https://docs.cloudmcp.run/blog/cloud-mcp-router
Stop overloading your agents with hundreds of tools. Cloud MCP Router brings progressive discovery, auth, and collision‑safe proxying to the Model Context Protocol so you can scale to real, multi‑app workflows.
> **TL;DR** — Agents don’t fail because they can’t call tools; they fail because we dump *all the tools* into the prompt at once. **Cloud MCP Router** adds a progressive, discovery‑driven layer in front of your MCP servers so the model sees *only what it needs, when it needs it*. Smaller prompts, fewer mis‑selections, and the ability to scale past arbitrary tool caps—without rewriting your existing MCP servers.
***
## The problem we’re solving
Modern agents juggle calendars, email, docs, CRMs, repos, issues, and PRs. Hand the model 80–200 tools at once and you get:
* **Context overload:** Tool schemas crowd out user content and inflate token cost.
* **Decision paralysis:** Long, look‑alike tool lists cause mis‑selections and retries.
* **Arbitrary caps:** Teams hide tools just to keep prompts short—shrinking what agents can actually do.
This pattern is especially acute in **Model Context Protocol (MCP)** ecosystems designed for rich, tool‑heavy workflows. The result: your “power user” assistant never reaches its potential because the tool layer isn’t designed for discovery.
***
## Meet Cloud MCP Router
**Cloud MCP Router** is a progressive, discovery‑driven layer that sits in front of your existing MCP servers (official, custom, or community) and turns any large toolset into a *query‑as‑you‑go* catalog. Instead of spraying every tool and schema into the base prompt, it exposes a small set of **router tools** that guide the agent through staged narrowing—only surfacing schemas at the moment of execution.
### What the model sees
* **Start with discovery:** `discover_server_actions` returns just the *relevant* actions (by intent), not entire schemas.
* **Drill in on demand:** `get_action_details` reveals parameters only for the chosen action.
* **Execute confidently:** `execute_action` runs with the now‑known parameters (with elicitation automatically bridged).
* **Stay unblocked:** `search_documentation` fetches the smallest useful doc snippets.
***
## How it works (under the hood)
### 1) Collision‑safe proxying of *any* MCP server
Point the Router at remote MCP servers (stdio) and it will connect, list their tools/prompts/resources, and **proxy** them through a clean namespace. It sanitizes and uniquifies names, enforces length limits, and pre‑checks for collisions before exposing anything to the client—so your model never sees confusing duplicates.
> **Why that matters:** Tools stay stable and human‑readable, even across many vendors and teams. No more “mystery collisions” that silently break calls.
### 2) **Router‑only mode** to shrink the prompt on demand
Flip **router‑only mode** and the Router removes all proxied tools from the surface area, leaving only the small, discovery‑first router tools. When you’re ready, you can selectively re‑enable a subset of actions or whole servers—and still avoid overload. Toggle via the `Router-Mode` switch on your dashboard.
### 3) Progressive discovery & execution API (the “router tools”)
* `discover_server_actions` — intent → relevant actions (fast, ranked).
* `get_action_details` — reveal schema *only* for the chosen action.
* `execute_action` — call the remote tool; **elicitation** is automatically bridged back to the upstream client session so the agent can ask follow‑ups mid‑execution without losing context.
### 4) Resources & prompts, too
Beyond tools, the Router proxies **resources** (with stable `proxy://…` URIs) and **prompts**, keeping names safe and avoiding collisions across servers. It even installs a catch‑all resource template so ad‑hoc reads “just work.”
### 5) Auth that fits your stack
The Router protects its MCP endpoint with **OAuth2** authentication against your Cloud MCP endpoint. It also publishes **well‑known protected resource metadata** so compliant clients know how to authorize. You get modern headers and correct `WWW-Authenticate` challenges out of the box.
***
## Why customers adopt Cloud MCP Router
* **Smaller prompts, lower cost:** Only a tiny subset of metadata is surfaced at any moment.
* **Fewer mistakes:** Staged, ranked choices beat a flat list of 100+ tools.
* **Scale past arbitrary caps:** Because schemas are fetched lazily, catalogs can grow to *hundreds* of actions without overwhelming the model.
* **Keep your stack:** No rewrites required—just add your existing MCP servers and go.
* **Built‑in resilience:** Doc search and auth‑recovery tools unstick agents without inflating the base prompt.
* **Operational sanity:** Health checks, safe namespacing, and collision detection remove brittle edges before they reach production.
***
## Example: What the agent’s loop looks like
1. **Discover:** “Open a PR that closes issue #123” → Router returns relevant actions like `repos.create_pull_request`, `issues.update`, ranked by intent.
2. **Detail:** Model asks details for `repos.create_pull_request`; Router returns just that action’s schema.
3. **Execute:** Model calls `execute_action` with the filled parameters; if extra info is needed mid‑call, elicitation is bridged back to the client session seamlessly.
**Net effect:** The base prompt stays lean. The agent makes better choices. Your catalog can grow without fear.
***
## Security & enterprise readiness
* **Session‑verified access tokens** (via Cloud MCP).
* **Correct `WWW-Authenticate` challenges** and **.well‑known metadata** for resource authorization.
***
## Ready to route?
Cloud MCP Router is available now. If you’re building assistants that need to span dozens of apps without cratering reliability or cost, this is the missing layer.
* **Try it with your existing MCP servers**—no rewrites.
* **Book a demo** to see progressive discovery reduce errors on your own workflows.
* **Talk to us** about rollout, SSO, and enterprise controls.
Let’s help your agents do more—with less.
# Deploy Remote MCP Servers from NPM, PyPI, or GitHub with Custom Deployments
Source: https://docs.cloudmcp.run/blog/deploy-any-mcp-server
Learn how to deploy any MCP server from NPM, PyPI, or GitHub using CloudMCP's new custom deployment feature. Support for private packages, real-time validation, and multiple package managers.
# Deploy Any MCP Server with CloudMCP's Custom Deployment Feature
CloudMCP now lets you deploy **any** MCP (Model Context Protocol) server, not just those in our curated registry. Whether your server is published on NPM, PyPI, or hosted on GitHub, you can deploy it with just a few clicks using our new custom deployment feature.
## Prerequisites
Before you begin, ensure you have:
* A cloudmcp.run account (sign up with GitHub at [cloudmcp.run](https://cloudmcp.run))
* The package name or GitHub URL of the MCP server you want to deploy
* Any required API keys or environment variables for your server
* (Coming Soon!) Authentication tokens for private packages
## What's New?
The custom deployment feature enables:
* **Universal package support**: Deploy from NPM, PyPI, or GitHub repositories
* **Private package deployment**: Use authentication tokens for private resources
* **Real-time validation**: Instant package verification as you type
* **Smart detection**: Automatic recognition of package vs repository inputs
* **Flexible configuration**: Full control over arguments and environment variables
## Step 1: Access Custom Deployment
Navigate to your CloudMCP dashboard and open the **Deployments** page. Scroll down to find the new "Custom Deployment" section.
Click the **"Deploy Custom Server"** button to open the deployment configuration dialog.
## Step 2: Configure Your Package Manager
The custom deployment dialog opens with a clean interface for configuring your server.
### Select Your Package Manager
Choose from four popular package managers based on your server's language:
* **NPX** - For Node.js packages (most MCP servers)
* **UVX** - For Python packages using UV
* **PIPX** - For traditional Python packages
* **BUNX** - For Bun runtime packages
The package manager determines how CloudMCP will install and run your server.
## Step 3: Enter Your Package or Repository
### Smart Input Detection
CloudMCP automatically detects what type of resource you're entering:
* **Package name**: Shows a package icon (📦)
* NPM: `@modelcontextprotocol/server-filesystem`
* PyPI: `mcp-server-sqlite`
* **GitHub repository**: Shows a GitHub icon (🐙)
* `https://github.com/owner/repo`
* `github.com/owner/repo`
### Real-time Validation
As you type, CloudMCP validates your input in real-time:
The validation checks:
* Package exists in the selected registry
* Latest version available
* Whether authentication is required
* Package metadata and description
> 💡 **Tip**: The validation happens automatically after you stop typing for 500ms, preventing unnecessary API calls while you type.
## Step 5: Add Configuration
### Deployment Name
Give your deployment a descriptive name to identify it in your dashboard.
Good naming examples:
* `production-database-server`
* `dev-filesystem-access`
* `customer-api-integration`
### Command Arguments (Optional)
Add any additional command-line arguments your server needs:
Common use cases:
* Specify configuration files: `--config /path/to/config.json`
* Set operation modes: `--mode production`
* Enable features: `--enable-feature-x`
### Environment Variables
Configure environment variables through the intuitive interface:
To add variables:
1. Click **"Add Variable"**
2. Enter the key (e.g., `API_KEY`)
3. Enter the value
4. Add more as needed
5. Remove unwanted variables with the trash icon
> 🔒 **Security Tip**: CloudMCP blocks potentially dangerous environment variables like `LD_PRELOAD` and `NODE_OPTIONS` for security.
## Step 6: Deploy Your Server
Once everything is configured, click **"Deploy Server"** to launch your deployment.
CloudMCP will:
1. Create a dedicated instance for your server
2. Install the package from your selected registry
3. Apply all environment variables securely
4. Start the server with your specified arguments
5. Provide you with a unique endpoint URL
## Step 7: Monitor Your Deployment
After deployment, you'll see your custom server in the Active Deployments section:
### Custom Deployment Indicators
Custom deployments are easy to identify:
* **"Custom" badge**: Distinguishes from registry servers
* **Package info**: Shows the exact package or repository deployed
* **Status indicator**: Real-time deployment status
* **Server URL**: Your unique endpoint for connecting
### Available Actions
* **Copy URL**: Get your server endpoint
* **Install in VSCode**: One-click VSCode integration
* **Stop**: Shut down when no longer needed
## Trial vs Paid Deployments
CloudMCP offers flexible deployment options:
### Trial Deployments
* **Duration**: 48 hours
* **Perfect for**: Testing and development
* **Limit**: Based on your account type
* **Automatic expiry**: Servers stop after 48 hours
### Paid Deployments
* **Duration**: Unlimited
* **Perfect for**: Production use
* **Resources**: Dedicated instances
* **Support**: Priority support included
## Best Practices
### 1. Validate Before Deploying
Let the real-time validation complete before clicking deploy. This catches issues early and saves time.
### 2. Use Descriptive Names
Help your future self by using clear, descriptive deployment names:
* ❌ `server1`
* ✅ `production-github-integration`
### 3. Secure Your Tokens
* Never share authentication tokens
* Rotate tokens regularly
* Use minimal required permissions
### 4. Test with Trials First
Use trial deployments to:
* Test server functionality
* Verify environment variables
* Check performance requirements
## Troubleshooting
### Validation Fails
**Package not found error:**
* Verify the package name spelling
* Check the package exists in the selected registry
* For GitHub, ensure the URL format is correct
**Version not available:**
* Check if the version exists
* Try using `latest` or no version
* For GitHub, verify the branch name
### Deployment Failures
**Environment variable errors:**
* Check for typos in variable names
* Verify required variables are set
* Ensure values don't contain invalid characters
**Resource limits:**
* Check your deployment quota
* [Upgrade](https://cloudmcp.run/pricing) plan if needed
* Delete unused deployments
### Connection Problems
**Can't connect to deployed server:**
* Verify the server started successfully
* [Redeploy](https://cloudmcp.run/dashboard/deployments) your server.
* Ensure your client supports the server's MCP version of the package.
* Send us a support
### Package Versions
Specify exact versions for stability:
```
NPM: @package/name@1.2.3
PyPI: package-name==1.2.3
```
### Multiple Configurations
Deploy the same server multiple times:
* Different environment variables for dev/prod
* Separate instances for different projects
* Isolated deployments for testing
### Monitoring Deployments
Track your custom deployments:
* Check status indicators regularly
* Monitor resource usage (coming soon)
* Set up alerts for failures (coming soon)
## What's Next?
Now that you've mastered custom deployments:
* 🚀 **Deploy your servers**: Start with your own MCP servers [--> Dashboard](https://cloudmcp.run/dashboard)
* 📦 **Try different registries**: Explore NPM, PyPI, and GitHub options
## Coming Soon
We're constantly improving custom deployments:
* **Deployment templates**: Save configurations for reuse
* **Bulk deployments**: Deploy multiple servers at once
* **Version management**: Easy updates and rollbacks
* **Performance metrics**: Detailed resource monitoring
## Need Help?
If you encounter any issues:
* Check our [blog](https://cloudmcp.run/blog) for updates and tutorials
* Join our [Discord community](https://discord.gg/cloudmcp)
* Contact me with the [support](https://cloudmcp.run/support) form!
***
*Ready to deploy your custom MCP server? [Start deploying →](https://cloudmcp.run/dashboard/deployments)*
# MCP Registry Launch: What the Official MCP Server Directory Means for Cloud MCP
Source: https://docs.cloudmcp.run/blog/mcp-registry-launch
Learn what the new official MCP Registry is, why it matters, and how Cloud MCP is integrating it to deliver an always-updated 'Official Servers' catalog and enterprise sub-registry support.
# MCP Registry Launch: the “single source of truth” moment for Model Context Protocol
**Published:** September 2025
In the last few days, the Model Context Protocol (MCP) team unveiled the **official MCP Registry** in preview-a canonical, open catalog and API for discovering publicly available MCP servers. It standardizes how servers are published and found, and introduces a federated model so organizations can run **public** or **private sub‑registries** on top of a shared upstream dataset. In short: one place to publish, many places to consume. ([mcp blog][1])
Why the fuss? Because MCP adoption has been gated by *discovery*-too many scattered lists, too many bespoke installers. The new registry provides that missing backbone: a public endpoint (with OpenAPI docs) that client authors and marketplaces can ingest, filter, and extend. The preview launch is explicit that breaking changes are still possible, but the direction is clear: a community‑maintained, vendor‑neutral index for MCP servers. ([GitHub][2])
## What’s new (and why it matters)
* **Canonical dataset & API.** A read‑only API at `registry.modelcontextprotocol.io` exposes server listings (e.g., `GET /v0/servers`), making it trivial for clients and aggregators to keep their catalogs fresh. ([techcommunity.microsoft.com][3])
* **Federated sub‑registries.** The official registry is the upstream; downstream *public* marketplaces (for specific clients) and *private* enterprise registries can mirror and enrich entries while sharing schemas and tooling. ([mcp blog][1])
* **Standard metadata (`server.json`).** Servers publish a compact manifest that links to actual packages (npm, PyPI, Docker, etc.), so the registry acts as a *metaregistry*-pointing to code hosted elsewhere while unifying discovery. ([GitHub][2])
## The conversation online (last 30 days)
Coverage and commentary have been brisk. Visual Studio Magazine framed the timing alongside Microsoft’s new “Awesome Copilot” MCP server, noting how concrete server implementations and a shared registry “arrived within days of each other.” That pairing makes MCP more usable inside familiar developer tools while standardizing how integrations are found. ([Visual Studio Magazine][4])
On the community side, explainer posts and how‑tos popped up within hours-from a Microsoft Tech Community walkthrough of the API endpoints to hands‑on guides for adding servers-underscoring immediate interest from practitioners. ([techcommunity.microsoft.com][3])
Developers are also debating ergonomics and governance on social platforms. A Hacker News thread (2 days old) discusses preview‑only API access, asks about a UI roadmap, and argues for value‑add curation (auth, vetting, RBAC) on top of the upstream feed-precisely the kind of healthy pressure that drives ecosystem polish. ([Hacker News][5])
***
## What the MCP Registry means for **Cloud MCP**
Cloud MCP exists to make **remote MCP servers** dead simple: paste a URL, OAuth to connect, and give your AI real capabilities across **mobile, web, and desktop**-no local binaries, no terminal. We already support one‑click deployments and **custom deployments** from npm/PyPI/GitHub with real‑time validation. The official registry lets us supercharge that experience. ([Cloud MCP][6])
Here’s how we’re integrating it:
1. **An “Official Servers” lane in *Find Servers*.** We’ll mirror the upstream registry on a frequent cadence and surface an **Official** filter inside your Cloud MCP dashboard, so you can browse the freshest list of published servers without hunting across the web. Entries remain install‑ready-click *Deploy* to spin up a remote instance with your environment variables and OAuth. ([Cloud MCP][7])
2. **Rich metadata from `server.json`.** We’ll ingest the registry’s standard metadata (name/namespace, packages, versions) to power **version pinning**, **changelogs**, and **install provenance** in Cloud MCP. You’ll see the source registry (npm/PyPI/Docker) and the exact package version you’re deploying. ([GitHub][2])
3. **Health & verification signals.** The official registry supports community moderation and deny‑listing. We’ll surface those signals alongside Cloud MCP’s own checks to help you choose trustworthy servers for production. ([mcp blog][1])
4. **Enterprise catalogs via sub‑registries.** For teams that curate internal tools, Cloud MCP will respect **private sub‑registries**: point Cloud MCP at your internal feed (mirrored from the official upstream), enforce org policy, and still benefit from the shared schema and ecosystem tooling. ([mcp blog][1])
**Bottom line for Cloud MCP users:** the official registry gives you a continuously updated, standardized source of MCP servers; Cloud MCP turns those listings into secure, remote deployments in seconds-**across every MCP‑capable client**. Discover centrally, deploy instantly, run anywhere. ([Cloud MCP][6])
***
### Sources & further reading
* **Official announcement:** “Introducing the MCP Registry” (preview, federated sub‑registries, OpenAPI). ([mcp blog][1])
* **Repo overview:** Registry README (preview status, maintainers, `server.json` model, live API docs). ([GitHub][2])
* **API how‑to:** Microsoft Tech Community (endpoints and curl usage). ([techcommunity.microsoft.com][3])
* **Ecosystem reaction:** Visual Studio Magazine (context and timing with Microsoft’s server). ([Visual Studio Magazine][4])
* **Community debate:** Hacker News thread on the registry launch (preview/API, curation, UI). ([Hacker News][5])
* **Cloud MCP background:** Deploy Remote MCP servers with OAuth; quick deploy and custom deployments from npm/PyPI/GitHub. ([Cloud MCP][6])
*If you’re ready to try it, head to your Cloud MCP dashboard, open **Find Servers**, and-soon-toggle **Official** to browse the upstream catalog, then **Deploy** to give your AI new superpowers in seconds.* ([Cloud MCP][7])
[1]: https://blog.modelcontextprotocol.io/posts/2025-09-08-mcp-registry-preview/ "Introducing the MCP Registry | mcp blog"
[2]: https://github.com/modelcontextprotocol/registry "GitHub - modelcontextprotocol/registry: A community driven registry service for Model Context Protocol (MCP) servers."
[3]: https://techcommunity.microsoft.com/discussions/appsonazure/how-to-use-the-newly-launched-mcp-registry/4452855 "How to use the newly launched MCP Registry | Microsoft Community Hub"
[4]: https://visualstudiomagazine.com/articles/2025/09/09/microsofts-awesome-copilot-mcp-server-joined-by-mcp-registry.aspx "Microsoft's 'Awesome Copilot MCP Server' Joined by MCP Registry -- Visual Studio Magazine"
[5]: https://news.ycombinator.com/item?id=45176580 "The MCP Registry | Hacker News"
[6]: https://cloudmcp.run/ "Cloud MCP - Remote Model Context Protocol Servers"
[7]: https://cloudmcp.run/blog/cloud-mcp-deployment-guide "How to Remotely Deploy an MCP Server (on cloudmcp.run)"
# Why Remote MCP Servers Unlock a New Era for AI Agents Beyond Desktop
Source: https://docs.cloudmcp.run/blog/remote-mcp-servers-mobile-ai-agents
The Model Context Protocol is transformative, but its desktop-only limitation leaves billions of mobile and web users behind. Remote MCP deployment changes everything - here's why it matters and what it enables.
# Why Remote MCP Servers Unlock a New Era for AI Agents Beyond Desktop
The Model Context Protocol (MCP) has revolutionized how AI agents interact with tools and data. But there's a critical limitation that's holding back its full potential: **traditional MCP servers only run on desktop**.
This means that the billions of users on mobile devices, web browsers, and other platforms are locked out of the AI agent revolution. They can chat with AI, but they can't give it real capabilities.
**Remote MCP deployment changes everything.**
## The Desktop Bottleneck
Traditional MCP servers use STDIO (standard input/output) for communication. This works great on a desktop where you have:
* Full file system access
* Ability to install native binaries
* Unrestricted process execution
* Generous CPU and memory resources
* Persistent background processes
But try running that on an iPhone, Android device, or in a web browser. You immediately hit walls:
* **Mobile sandboxing** prevents file system access
* **App store restrictions** block binary installations
* **Limited resources** can't handle heavy computation
* **No background processes** means no persistent services
* **Security models** prevent system-level operations
The result? Mobile and web users get a watered-down AI experience. Their AI assistants can talk, but they can't *do*.
## Enter Remote MCP Servers
Remote MCP servers flip the script. Instead of running locally, they run in the cloud and communicate over HTTP with OAuth authentication. This simple architectural change unlocks profound capabilities.
Suddenly, your mobile AI assistant can:
* **Manage infrastructure** on AWS, Azure, or Google Cloud
* **Query databases** like PostgreSQL, MongoDB, or BigQuery
* **Execute code** in sandboxed Python or Node.js environments
* **Automate browsers** with Puppeteer or Playwright
* **Process documents** with OCR and AI extraction
* **Control Docker containers** and Kubernetes clusters
* **Access file systems** for reading and writing documents
* **Schedule tasks** with cron-like automation
None of this is possible with local MCP servers on mobile. All of it becomes trivial with remote deployment.
## Real-World Impact: Use Cases That Matter
### **Mobile DevOps on the Go**
Imagine you're a DevOps engineer getting paged at 2 AM. Instead of rushing to your laptop, you grab your phone and tell your AI assistant: "Check the Kubernetes cluster health and rollback the last deployment if CPU usage is above 80%."
With remote MCP servers for Kubernetes and cloud providers, your mobile AI becomes a full DevOps command center.
### **Field Data Collection and Processing**
A field researcher collecting samples can use their phone to:
* Upload photos to cloud storage
* Run image analysis with specialized tools
* Update databases with findings
* Generate reports with complex formatting
* All through natural language commands
The heavy lifting happens on remote servers while the phone provides the interface.
### **Executive Decision Support Anywhere**
A CEO traveling can ask their mobile AI: "Pull yesterday's sales data from BigQuery, compare it to last quarter's average, and create a presentation for the board meeting."
Remote MCP servers for BigQuery, data visualization, and document generation make this possible without touching a laptop.
### **Creative Workflows Unleashed**
Content creators can leverage remote MCP servers for:
* **FFmpeg servers** for video processing and conversion
* **ImageMagick servers** for batch image editing
* **Pandoc servers** for document format conversion
* **Git servers** for version control and collaboration
These tools require binaries and processing power that mobile devices simply don't have.
## The Technical Advantages
### **1. Resource Liberation**
Heavy computational tasks run on powerful cloud servers, not battery-constrained mobile devices. Your phone becomes a lightweight controller for heavyweight capabilities.
### **2. Universal Access**
The same MCP servers work across all platforms - iOS, Android, web browsers, smart TVs, even voice assistants. Write once, deploy everywhere.
### **3. Team Collaboration**
Remote servers can be shared across teams. Everyone accesses the same tools with the same configurations, ensuring consistency and enabling collaboration.
### **4. Security and Compliance**
OAuth authentication, encrypted connections, and cloud-grade security protect sensitive operations. Audit logs track every action for compliance.
### **5. Always Up-to-Date**
Remote servers can be updated instantly without requiring app store approvals or user updates. Bug fixes and new features deploy immediately to all users.
## Breaking Down Specific Server Categories
### **Development and Code Execution**
* **Code sandboxes** (Python, Node.js, Ruby) for running scripts
* **Database clients** for querying and managing data
* **Git operations** for repository management
* **Container orchestration** for Docker and Kubernetes
*Why they need remote deployment:* Require runtimes, compilers, and system access unavailable on mobile.
### **Data Processing and Analytics**
* **BigQuery, Snowflake, Databricks** for enterprise analytics
* **Apache Spark** for distributed computing
* **ETL pipelines** for data transformation
* **Machine learning frameworks** for model training
*Why they need remote deployment:* Massive computational requirements and specialized libraries.
### **Automation and Integration**
* **Browser automation** with Puppeteer/Playwright
* **API orchestration** for complex workflows
* **Webhook handlers** for event-driven automation
* **Scheduled tasks** for recurring operations
*Why they need remote deployment:* Need persistent processes and full browser environments.
### **Media and Content**
* **FFmpeg** for video/audio processing
* **ImageMagick** for image manipulation
* **Pandoc** for document conversion
* **PDF generation** and manipulation
*Why they need remote deployment:* Require native binaries and significant processing power.
## The Mobile-First AI Future
We're entering an era where AI agents aren't confined to desktop applications. With remote MCP servers, every device becomes a portal to unlimited AI capabilities.
Consider the implications:
* **Systemprompt** and similar mobile apps can offer the same power as desktop AI tools
* **Voice assistants** like Alexa or Google Assistant could integrate MCP capabilities
* **Smartwatches** could control complex infrastructure with a tap
* **AR glasses** could overlay AI-powered tools onto the physical world
This isn't science fiction - it's happening now. Companies like Cloudflare are building infrastructure for remote MCP servers. Platforms like **CloudMCP.run** are making deployment as simple as clicking a button.
## Getting Started with Remote MCP
The transition from desktop-only to remote MCP is straightforward:
1. **Choose your servers**: Identify which MCP servers would benefit your users
2. **Deploy remotely**: Use platforms like CloudMCP.run for instant deployment
3. **Configure authentication**: Set up OAuth for secure access
4. **Connect clients**: Mobile apps and web interfaces can now access your servers
5. **Monitor and iterate**: Track usage, gather feedback, and expand capabilities
## The Bottom Line
Remote MCP servers aren't just a technical evolution - they're a paradigm shift. They transform AI agents from desktop-bound assistants into ubiquitous, powerful tools accessible from any device, anywhere.
The companies and developers who recognize this shift early will have a massive advantage. They'll be able to offer AI experiences that their desktop-only competitors simply can't match.
The future of AI isn't sitting at a desk. It's in your pocket, on your wrist, in your car, and everywhere else you go. Remote MCP servers are the bridge to that future.
**Ready to deploy your first remote MCP server?** [Get started with CloudMCP.run →](https://cloudmcp.run)
***
*CloudMCP.run makes it simple to deploy any MCP server remotely with OAuth security, instant scaling, and pay-per-use pricing. No infrastructure headaches, just pure innovation.*
# Introducing CloudMCP.run - Deploy MCP Servers Without the Headaches
Source: https://docs.cloudmcp.run/blog/welcome
We're excited to introduce CloudMCP.run, a platform that makes deploying Model Context Protocol servers as easy as clicking a button.
# Introducing CloudMCP.run - Deploy MCP Servers (Remotely) Without the Headaches
If you've been working with AI agents and the Model Context Protocol (MCP), you know the drill. You've got a brilliant idea for an MCP server that could give AI models access to new capabilities - maybe it's a custom database connector, a specialized API integration, or a unique tool that could supercharge AI workflows.
But then reality hits.
Setting up the infrastructure. Managing deployments. Configuring security. Handling scaling. Monitoring performance. Before you know it, you're spending more time wrestling with DevOps than actually building your innovation.
**That's why we built CloudMCP.run.**
## What is CloudMCP.run?
CloudMCP.run is a platform that lets you deploy Model Context Protocol servers instantly. We handle all the infrastructure complexity so you can focus on what matters - building amazing AI tools and integrations.
Think of it as Vercel or Netlify, but specifically designed for MCP servers. With just a few clicks, you can have a production-ready MCP server running in the cloud, complete with:
* 🚀 **Instant deployment** - Launch pre-configured MCP servers in minutes
* 📈 **Auto-scaling** - Your servers scale automatically based on demand
* 🔒 **Built-in security** - Robust authentication and authorization out of the box
* ⚡ **Optimized performance** - Low-latency responses for real-time AI applications
* 🔧 **GitHub integration** - Sign up and manage everything with your GitHub account
## Why MCP Servers Need Better Infrastructure
The Model Context Protocol is revolutionizing how AI models interact with external tools and data sources. By providing a standardized way for AI to access capabilities beyond their training data, MCP opens up incredible possibilities.
But running MCP servers effectively requires:
* **High availability** - AI agents need reliable access to tools
* **Low latency** - Real-time applications can't wait for slow responses
* **Security** - Protecting sensitive data and preventing unauthorized access
* **Scalability** - Usage can spike unexpectedly as AI applications grow
Managing all of this yourself is a massive undertaking that distracts from your core innovation.
## How CloudMCP.run Works
We've made deployment dead simple:
1. **Choose your server** - Pick from our library of popular MCP servers or bring your own
2. **Configure** - Set up your environment variables and authentication
3. **Deploy** - Click deploy and your server is live in minutes
4. **Connect** - Use your server URL in any MCP-compatible AI application
That's it. No Kubernetes configs. No Docker orchestration. No sleepless nights debugging infrastructure issues.
## What's Available Today
We're launching with support for the most popular MCP server configurations, including:
* Database connectors (PostgreSQL, MySQL, MongoDB)
* API integrations (Slack, GitHub, custom REST APIs)
* File system tools
* Custom computation servers
* And many more...
Each server comes pre-configured with best practices for security and performance. Just add your credentials and deploy.
## Our Vision
We believe the future of AI lies in its ability to interact with the world through tools and integrations. MCP is the protocol that makes this possible, but it shouldn't require a PhD in cloud architecture to deploy.
Our mission is to democratize access to MCP infrastructure, allowing developers of all skill levels to contribute to the AI ecosystem. Whether you're a solo developer with a clever idea or a team building enterprise AI solutions, CloudMCP.run gives you the foundation to move fast and build confidently.
## Get Started Today
Ready to deploy your first MCP server? Here's how to get started:
1. **Sign up** at [cloudmcp.run](https://cloudmcp.run) using your GitHub account
2. **Start your free trial** - No credit card required
3. **Deploy your first server** - Be up and running in under 5 minutes
We're offering a generous free tier so you can experiment and build without worrying about costs. When you're ready to scale, our transparent pricing grows with your usage.
## What's Next
This is just the beginning. We're working on exciting features including:
* **Custom domains** for your MCP servers
* **Advanced monitoring** and debugging tools
* **Team collaboration** features
* **Private server registry** for proprietary MCP implementations
* **One-click templates** for common use cases
## Join Us
The AI revolution needs better infrastructure. If you're tired of fighting with deployments when you should be building the future, give CloudMCP.run a try.
Have questions? Feedback? Want to request a specific MCP server? We'd love to hear from you. Drop us a line at [hello@cloudmcp.run](mailto:hello@cloudmcp.run) or join our [Discord community](#).
Let's build the future of AI tooling together - without the infrastructure headaches.
***
*Ready to deploy your first MCP server? [Get started free at cloudmcp.run →](https://cloudmcp.run)*
# Development
Source: https://docs.cloudmcp.run/development
Preview changes locally to update your docs
## Code formatting
We suggest using extensions on your IDE to recognize and format MDX. If you're a VSCode user, consider the [MDX VSCode extension](https://marketplace.visualstudio.com/items?itemName=unifiedjs.vscode-mdx) for syntax highlighting, and [Prettier](https://marketplace.visualstudio.com/items?itemName=esbenp.prettier-vscode) for code formatting.
## Troubleshooting
## Image
### Using Markdown
The [markdown syntax](https://www.markdownguide.org/basic-syntax/#images) lets you add images using the following code
```md theme={null}

```
Note that the image file size must be less than 5MB. Otherwise, we recommend hosting on a service like [Cloudinary](https://cloudinary.com/) or [S3](https://aws.amazon.com/s3/). You can then use that URL and embed.
### Using embeds
To get more customizability with images, you can also use [embeds](/writing-content/embed) to add images
```html theme={null}
```
## Embeds and HTML elements
... snippet content ...