Skip to main content

Quick Start Guide

Connect your first MCP server to ChatGPT, Claude, Cursor, or any MCP-compatible AI in just a few minutes.

Time to complete: 5–10 minutes

What You'll Accomplish​

By the end of this guide, you'll have:

  • An MCP server enabled in your workspace
  • Required provider connections connected, verified, and working
  • Server showing Ready (for example All connections ready)
  • Your MCP server connected to your AI assistant
  • Successfully called a tool from your AI

What is MCP? (60 seconds)​

MCP (Model Context Protocol) is a standard that lets AI tools discover and call external capabilities.

Instead of your AI being limited to what it was trained on, MCP lets it:

  • Access real-time data from APIs
  • Perform actions in external services
  • Extend its capabilities dynamically

An MCP server is one URL that exposes a related set of tools. Add the URL to your AI, and every tool on that server becomes available.

Example: A marketing MCP server might give your AI tools to list email campaigns, check replies, generate copy, and create contacts—all from one URL.

Prerequisites​

Before starting:

  • An MCPBundles account (Sign up free)
  • An AI tool that supports MCP (ChatGPT, Claude, Cursor, VS Code, etc.)
  • Credentials for any providers the MCP server uses (API keys or OAuth)

Step 1: Choose or Connect an MCP Server​

Option A: Connect a server from the directory​

  1. Go to MCP Servers
  2. Use the unified list with segment filters All, In workspace, and Directory (plus search and directory chips)
  3. Open a server that fits your needs
  4. Click Connect on the row or detail page

The server is in your workspace. If providers still need connections, the header shows Needs setup (for example Connect N connections to use this MCP server) until Step 2 is done.

Option B: Build a private MCP server​

There is no top-level Create button on the MCP Servers list. The in-app entry point on that page today is Add remote MCP URL (header action on MCP Servers) for an external MCP endpoint you host elsewhere.

For MCPBundles-hosted private or composed servers, follow Creating Custom MCP Servers and Composing MCP Servers. Connect providers in the Connections section on each server you use.

See Creating Custom MCP Servers and Composing MCP Servers.

New to MCP servers? See Understanding MCP Servers for how they work.

Step 2: Configure Connections​

Most MCP servers require connections (API keys or OAuth) to access external APIs (Smartlead, GitHub, Slack, etc.). You'll add and verify these from your MCP server page in just a few clicks.

Team Workspaces

If you're in a team workspace (2+ members), credentials are shared with all members. When one person connects a provider, everyone in the workspace can use it — no duplicate setup needed. See Understanding Workspaces for details.

Check What's Needed​

  1. On your MCP server page, check the header status (same labels across list, detail, and Studio):

    • Ready — all required connections verified (All connections ready, or All connections ready — install in your AI host to start if you have not run a tool recently)
    • Needs setup — missing connections (for example Connect N connections to use this MCP server)
    • Needs attention — a connection exists but is unverified, expired, or missing OAuth scopes
    • Disabled — server not enabled in this workspace (message may still say Connect… or Enable this MCP server… depending on connection state)
    • Status unknown — readiness could not be resolved yet; refresh the page
    • Connection unavailable — a required sign-in path is temporarily unavailable for this server

  2. Scroll to the Connections section

  3. Each required provider shows whether a connection is bound and verified

Adding connections​

When a provider still needs setup:

  1. Click Connect on the connection row (labeled with the service name)
  2. A credential setup panel slides in from the right

Important: The credential panel stays open throughout the entire process — you won't lose your place.

For OAuth Providers (GitHub, Slack, Sentry, etc.)​

  1. Enter an optional credential name (e.g., "Production Account", "Work Slack")
  2. Click Connect with [Provider]
  3. OAuth window opens → Log in to the provider
  4. Review and approve the requested permissions
  5. OAuth window closes → You return to the credential panel
  6. When the provider supports it, choose a validation tool or run Validate Now from the connection panel

For API Key Providers (Smartlead, OpenAI, etc.)​

  1. Get your API key from the provider's website (usually Settings → API Keys)
  2. Paste the API key into the credential panel
  3. (Optional) Enter a credential name
  4. Click Create Connection (some flows still label this Create Credential)

API key connections often auto-validate on save. You can still pick a validation tool or use Validate Now when status is Unverified or Error.

Choose a Validation Tool​

After adding your credential, pick a tool marked Recommended — these are read-only and safe.

  1. Click Use as validator on a recommended tool
  2. Click the green Validate Now button
  3. Wait 1–3 seconds:
    • Success: Status changes to VERIFIED
    • Failed: Error message explains what went wrong

Confirm the Server is Ready​

When the header shows Ready, you're ready for Step 3.

Step 3: Add Your MCP Server to Your AI​

Before proceeding: Confirm Ready status in the header. If not, finish Step 2 first.

Get Your MCP Server URL​

  1. On your MCP server page, confirm Ready (tools need verified connections — the URL row can appear before every connection is verified)
  2. Find the setup card near the top with the MCP Server URL (visible when the server is enabled in your workspace and exposes at least one tool)
  3. Copy the URL — e.g. https://mcp.mcpbundles.com/bundle/your-server-slug
  4. Click Copy on the URL row (expand Setup instructions for client-specific steps)
MCP URL vs tool access

The URL copy row appears when the server is enabled in your workspace and has tools. You still need verified connections before tools work in Studio or your AI client — finish Step 2 if calls fail or the header is not Ready.

Add to Your AI Tool​

Choose your platform — see Integration Guides for full steps:

ChatGPT: Settings → Apps → Create app → paste MCP Server URL → OAuth → Connect

Cursor: Add to .cursor/mcp.json with your MCP server URL, then restart Cursor

Claude Desktop: Settings → Custom Connectors → paste URL

Claude Code (CLI):

claude mcp add --transport http my-server https://mcp.mcpbundles.com/bundle/your-server-slug

Step 4: Test Your Tools​

In Your AI​

  1. Open a new conversation
  2. Ask your AI to list available tools: 
    What MCP tools do you have available?
  3. Your AI should show tools from your MCP server
  4. Try calling one: 
    Use the list_campaigns tool to show my email campaigns

Try MCPBundles Studio First​

Before testing in your AI, you can run tools interactively in the dashboard:

→ MCPBundles Studio

If Tools Don't Appear​

Check in order:

  1. Server status shows Ready
  2. All required providers show Verified (green)
  3. Server is enabled on your account
  4. You restarted your AI client after adding the URL
  5. You're using the correct MCP URL from the server page setup card (/bundle/your-slug)

Still stuck? See Troubleshooting.

What's Next?​

Learn the Concepts​

Expand Your Setup​

Get Help​

Need Help?​

Congratulations! You're now using MCP servers to extend your AI with real-world capabilities.