MCP Marketplace: Browse 500+ Providers and 10,000+ AI Tools

Glama indexes 20,000+ MCP servers. Smithery has 8,000+. mcp.so has 6,000+. There's no shortage of servers to find.

The problem is everything that happens after you find one.

You pick a promising-looking Stripe MCP server from a directory. Now you need to clone the repo, install its dependencies (hope they don't conflict with yours), figure out whether it uses env or args for the API key, add your key to a JSON config file in plaintext, start the process, and configure your AI client to talk to localhost:3000. If you're lucky, it works. If the repo hasn't been updated in three months, it probably doesn't.

Repeat that for every service you want to connect. We got to five local MCP server processes before we gave up and built something better.

What we wanted a marketplace to be

Not a directory. Directories solve discovery, and the existing ones are good at it. What's missing is the step between "I found a server" and "my AI is using it."

An actual marketplace should:

• Let you connect your Stripe/HubSpot/Gmail/whatever account in a couple clicks
• Handle credentials properly (encrypted, not plaintext in a config file)
• Share access across your team without everyone setting up their own servers
• Keep working when your laptop sleeps, your dependencies update, or the upstream repo breaks
• Work with Claude, Cursor, Windsurf, ChatGPT — not just one client

That's what we built MCPBundles to do. 500+ providers, 10,000+ tools, all hosted. You browse, connect your account, and the tools are live.

How it actually works

You go to the providers page, find Stripe (or HubSpot, or Gmail, or PostgreSQL — whatever you need). Click connect. OAuth services open a browser window — authorize and you're done. API key services have a field. The credential gets encrypted and stored per-workspace.

Then you add the tools to a bundle. A bundle is one MCP endpoint that combines multiple services. So instead of your AI connecting to five separate servers, it connects to one URL and has Stripe + HubSpot + PostgreSQL + Gmail + Sentry in the same session. That's what makes prompts like "find the customer in Stripe, pull their HubSpot record, query the database for usage data, and draft an email" work in a single conversation.

Your AI client gets one URL:

{
  "mcpServers": {
    "my-tools": {
      "url": "https://mcp.mcpbundles.com/bundle/your-bundle"
    }
  }
}

Or use the CLI if you want runtime discovery and cross-service scripting:

pip install mcpbundles
mcpbundles connect my_workspace
mcpbundles init

Works with Claude Desktop, Claude Code, Cursor, Windsurf, ChatGPT, or anything that speaks MCP.

How MCPBundles compares to directories

This isn't a knock on directories — we use Glama and Smithery to discover open-source servers too. They serve a different purpose.

	MCPBundles	Glama	Smithery	mcp.so
Purpose	Connect and use	Discover and index	Discover and install	Discover and browse
Servers	500+ providers, all hosted	20,000+ indexed	8,000+ indexed	6,000+ indexed
Auth	Built-in OAuth + API key, encrypted	You manage	Some hosted auth	You manage
Team sharing	One workspace, everyone gets access	Per-user config	Per-user config	Per-user config
Cross-service	Bundle services into one endpoint	One server at a time	One server at a time	One server at a time
Hosting	Fully hosted	Self-hosted (mostly)	Mix	Self-hosted

The directories have wider coverage because they index everything, including weekend projects and abandoned repos. MCPBundles has fewer providers but every one of them works, is maintained, and has proper credential handling.

If you're exploring what's possible with MCP, start with a directory. If you need tools that work reliably in production with your team, that's what MCPBundles is for.

What people actually use it for

Based on our workspace data, the most-enabled providers across all teams:

• Stripe — by far. Customer lookups during debugging, subscription audits, invoice checks.
• PostgreSQL — ad-hoc queries against production data without leaving the AI conversation.
• HubSpot — contact search, deal stage checks, CRM data for customer support.
• GitHub — issue creation, PR review, repo exploration.
• Google Search Console — SEO checks, indexing status, traffic analysis.
• Gmail — inbox search, drafting replies, email automation workflows.

Most teams start with 2-3 services and expand as they realize how much time it saves. The jump from "I use one MCP server" to "I use five" happens fast once the setup friction disappears.

Browse the marketplace

• Providers — 500+ services by category
• Tools — search 10,000+ individual operations
• What is MCP? — if you're new to the protocol

FAQ

What is an MCP marketplace?

A platform where you find, connect, and use MCP servers — not just browse them. A marketplace handles hosting, authentication, and credential management so tools work immediately after you connect your account.

How many tools does MCPBundles have?

10,000+ individual tools across 500+ providers. CRM, payments, databases, email, analytics, SEO, dev tools, cloud infrastructure — every category a dev team needs.

Does it work with Claude and Cursor?

Yes — Claude Desktop, Claude Code, Cursor, Windsurf, ChatGPT, and any MCP-compatible client. Standard protocol, one URL.