Connecting Servers
Once you've created an MCP server and added tools, you need to connect AI assistants to it so they can discover and use your tools. This guide explains how to find your connection information and connect different AI assistants to your MCP servers.
What You Need to Connect
To connect an AI assistant to your MCP server, you need two pieces of information:
Server URL - The endpoint where your MCP server is accessible. This is a unique URL provided by MCP Functions that points to your specific server.
API Key - The authentication key for your server. This is the API key you created when setting up the MCP server. It ensures that only authorized AI assistants can connect to your tools.
Both of these are provided by MCP Functions when you create an MCP server. You'll need to copy them and provide them to the AI assistant you want to connect.
Why These Are Required
These two pieces of information serve important purposes:
Server URL - Tells the AI assistant where to find your server. Without this, the AI doesn't know where to connect.
API Key - Authenticates the connection. This ensures that only people who have the API key can connect to your server, protecting your tools from unauthorized access.
Think of it like a house address (Server URL) and a key to the door (API Key). The address tells you where to go, and the key lets you in.
Finding Your Connection Information
Your connection information is available in the MCP Functions dashboard. Here's how to find it:
Step-by-Step: Finding Connection Details
Navigate to your MCP server - Go to your workspace and click on the MCP server you want to connect
Open server settings - Look for a "Settings", "Connection", or "Connection Info" section or button
View connection information - You'll see:
The Server URL (a URL like `https://api.mcpfunctions.com/mcp/your-server-id`) - The API Key (a long string of characters) - Connection instructions specific to your server
- **Copy the information** - Copy both the Server URL and API Key. You'll need to paste them into the AI assistant's configuration ### Important Security Notes
API Key is shown only once - When you first create an API key, it's displayed once. After that, you can't view it again. Make sure to copy and save it securely when it's first shown.
Treat API keys like passwords - Don't share API keys publicly, commit them to code repositories, or include them in screenshots. If an API key is compromised, revoke it immediately and create a new one.
Use different keys for different purposes - Consider creating separate API keys for development and production, or for different AI assistants. This allows you to revoke access for one without affecting others.
What the Connection Information Looks Like
Your connection information will look something like this:
Server URL: https://api.mcpfunctions.com/mcp/abc123def456 API Key: mcp_abc123xyz789def456ghi012jkl345mno678pqr901
The exact format may vary, but you'll always have a URL and an API key. Copy both of these - you'll need them in the next step.
General Connection Process
While each AI platform has its own interface for connecting MCP servers, the general process is similar across all platforms. Understanding this process helps you connect to any AI assistant.
The Standard Connection Steps
Most AI assistants follow this general process:
Open AI assistant settings - Navigate to the settings or configuration section of your AI assistant
Find MCP/Integration section - Look for sections labeled "MCP Servers", "Custom Tools", "Integrations", "Extensions", or similar
Add a new server connection - Click a button to add a new MCP server connection
Enter connection details - Provide the Server URL and API Key you copied from MCP Functions
Save the configuration - Save the connection settings
Test the connection - Verify that the AI assistant can connect and discover your tools
What Happens During Connection
When you save the connection information, the AI assistant will:
Connect to your server - Establish a connection using the Server URL
Authenticate - Provide the API key to authenticate the connection
Discover tools - Request a list of available tools from your server
Load tool information - Receive tool names, descriptions, and parameter schemas
Make tools available - Your tools are now available for the AI to use
This entire process happens automatically. If everything is configured correctly, your tools will be available immediately. If there's an issue (like an incorrect API key), you'll see an error message explaining what went wrong.
Verifying the Connection
After connecting, you should verify that everything works:
Check tool discovery - Ask the AI assistant to list available tools. It should show your tools.
Test a tool - Ask the AI a question that would require one of your tools. For example, if you have a weather tool, ask "What's the weather in New York?"
Verify results - Make sure the AI can successfully use your tools and return correct results
If the AI can list your tools and use them successfully, your connection is working correctly!
Connecting Specific AI Assistants
While the general process is similar, each AI platform has its own interface. Here's guidance for connecting to popular AI assistants:
Connecting to ChatGPT
ChatGPT supports MCP servers through its settings:
Open ChatGPT settings - Click on your profile or settings icon in ChatGPT
Navigate to Integrations or MCP - Look for "Integrations", "MCP Servers", or "Custom Tools" in the settings menu
Add MCP server - Click "Add Server" or "Connect Server"
Enter connection details:
Server URL: Paste your MCP server URL - API Key: Paste your API key - Server Name (optional): Give it a friendly name like "My Tools"
- **Save and test** - Save the configuration and test by asking ChatGPT to list available tools
Note: The exact interface may vary depending on your ChatGPT version and subscription plan. If you don't see MCP options, make sure you're using a version that supports MCP.
Connecting to Claude
Claude (Anthropic's AI assistant) supports MCP servers:
Open Claude settings - Access the settings or preferences in Claude
Find MCP configuration - Look for "MCP Servers", "Tools", or "Integrations"
Add server - Click to add a new MCP server connection
Provide connection info:
Enter your Server URL - Enter your API Key - Optionally provide a description
- **Connect** - Save the connection and verify Claude can see your tools
Testing: Ask Claude "What tools do you have available?" to verify it can see your MCP tools.
Connecting to Other AI Assistants
Other AI assistants that support MCP will have similar processes:
Look for settings related to "MCP", "Tools", "Integrations", or "Extensions"
Find an option to add or connect an MCP server
Provide the Server URL and API Key
Save and test the connection
If you're using an AI assistant that supports MCP but you're not sure how to connect, check the AI assistant's documentation for MCP or tool integration instructions.
Platform-Specific Notes
Some platforms require specific formats - Some AI assistants might require the connection information in a specific format (like a configuration file). Check the platform's documentation.
Connection limits - Some platforms may limit how many MCP servers you can connect. Check the platform's limits if you plan to connect multiple servers.
Version requirements - Make sure you're using a version of the AI assistant that supports MCP. Older versions might not have this feature.
Testing Your Connection
After connecting your AI assistant to your MCP server, it's important to test that everything works correctly. This helps you catch any issues early and ensures your tools are accessible.
Basic Connection Test
The simplest test is to ask the AI assistant to list available tools:
Ask: "What tools do you have available?" or "List your available tools"
Expected result: The AI should list all the tools from your MCP server, including their names and descriptions
If it works: Your connection is successful and the AI can discover your tools
If it doesn't work: Check the error message and verify your connection information is correct
Functional Test
Once you've verified tool discovery, test that the AI can actually use your tools:
Ask a question that requires your tool - For example, if you have a weather tool, ask "What's the weather in New York?"
Observe the AI's behavior - The AI should recognize that it needs to use your tool and call it automatically
Verify the result - Make sure the AI returns the correct result from your tool
Error Testing
It's also good to test error handling:
Test with invalid inputs - Try asking questions that would cause your tool to return errors
Verify error messages - Make sure errors are handled gracefully and the AI can explain what went wrong
What to Check If Connection Fails
If the connection doesn't work, check these common issues:
Server URL is correct - Make sure you copied the entire URL correctly, including https://
API Key is correct - Verify the API key matches exactly what's shown in MCP Functions
API Key hasn't been revoked - If you revoked the API key, you'll need to create a new one and update the connection
Server is active - Make sure your MCP server is active and has tools in it
Network connectivity - Ensure your device can reach the MCP Functions servers
AI assistant supports MCP - Verify you're using a version of the AI assistant that supports MCP
Common Error Messages
Here are some common error messages and what they mean:
"Connection failed" - The AI assistant couldn't reach your server. Check the Server URL and network connectivity.
"Authentication failed" - The API key is incorrect or has been revoked. Verify the API key and create a new one if needed.
"No tools found" - The connection worked, but your server has no active tools. Make sure you've created and deployed tools in your MCP server.
"Server not found" - The Server URL is incorrect or the server doesn't exist. Verify the URL in MCP Functions.
Managing Multiple Connections
You might want to connect the same MCP server to multiple AI assistants, or connect multiple MCP servers to one AI assistant. Here's how to manage these scenarios:
Connecting One Server to Multiple AI Assistants
You can connect the same MCP server to multiple AI assistants (like both ChatGPT and Claude):
Use the same connection information - The Server URL and API Key work with any MCP-compatible AI assistant
No conflicts - Multiple AI assistants can connect to the same server simultaneously without issues
Shared tools - All connected AI assistants will see the same tools
Independent usage - Each AI assistant uses the tools independently - they don't interfere with each other
This is useful if you want to use the same tools across different AI platforms or if team members use different AI assistants.
Connecting Multiple Servers to One AI Assistant
You can connect multiple MCP servers to a single AI assistant:
Add multiple connections - In the AI assistant's settings, add multiple MCP server connections
Each server has its own tools - The AI assistant will see tools from all connected servers
Organize by purpose - You might connect different servers for different purposes (e.g., one for customer data, one for analytics)
Tool names must be unique - If multiple servers have tools with the same name, the AI assistant will need to disambiguate or you should rename tools to be unique
This is useful for organizing tools by function or team, or for separating development and production tools.
Best Practices for Multiple Connections
Use descriptive server names - When adding servers to an AI assistant, give them descriptive names so you know which is which
Organize logically - Group related tools in the same server to make them easier to manage
Use different API keys - Consider using different API keys for different AI assistants so you can revoke access independently
Monitor usage - Check execution logs to see which AI assistants are using which tools
Updating Connection Information
Sometimes you need to update your connection information. Here's when and how to do it:
When to Update
You might need to update connection information when:
API key is compromised - If you suspect someone has your API key, revoke it and create a new one
Rotating keys for security - Periodically rotating API keys is a good security practice
Server URL changes - If your server URL changes (rare, but possible), you'll need to update it
Switching to a different API key - If you want to use a different API key for an AI assistant
How to Update
To update connection information in an AI assistant:
Open the AI assistant's settings
Find the MCP server connection you want to update
Edit the connection (usually there's an "Edit" or "Update" button)
Update the Server URL or API Key as needed
Save the changes
Test the connection to verify it still works
Revoking and Recreating API Keys
If you need to revoke an API key:
Go to your MCP server settings in MCP Functions
Find the API key you want to revoke
Click "Revoke" or "Delete"
Create a new API key if needed
Update all AI assistants that were using the old key with the new key
Important: When you revoke an API key, all AI assistants using that key will immediately lose access. Make sure to update them with the new key before revoking the old one, or accept that there will be a brief period where the connection is broken.
Troubleshooting Connection Issues
If you're having trouble connecting an AI assistant to your MCP server, here are common issues and solutions:
Connection Won't Establish
Symptoms: The AI assistant can't connect to your server at all
Possible causes and solutions:
Incorrect Server URL - Double-check that you copied the entire URL correctly, including https://
Network issues - Check your internet connection and try again
Server is down - Check MCP Functions status page or try again later
Firewall blocking - If you're on a corporate network, check if firewalls are blocking the connection
Authentication Fails
Symptoms: Connection works but authentication fails
Possible causes and solutions:
Incorrect API Key - Verify the API key matches exactly what's shown in MCP Functions
API Key was revoked - Check if the API key is still active in MCP Functions
Extra spaces or characters - Make sure you didn't accidentally include extra spaces when copying the API key
Wrong API key - Ensure you're using the API key for the correct MCP server
Tools Not Discovered
Symptoms: Connection works but AI assistant can't see your tools
Possible causes and solutions:
No tools in server - Make sure you've created and deployed tools in your MCP server
Tools are inactive - Check that your tools are set to "Active" status, not "Inactive"
Server not fully initialized - Wait a moment and try again, or refresh the connection
Cache issues - Some AI assistants cache tool lists; try disconnecting and reconnecting
Tools Don't Work
Symptoms: AI assistant can see tools but can't use them
Possible causes and solutions:
Tool errors - Check execution logs in MCP Functions to see if tools are failing
Missing secrets - Verify that required workspace secrets are configured
Tool code issues - Test your tools in MCP Functions to ensure they work correctly
Parameter issues - Check that the AI is providing parameters in the correct format
Getting Help
If you've tried these solutions and still have issues:
Check execution logs - Look at logs in MCP Functions to see what's happening
Test tools directly - Use the testing interface in MCP Functions to verify tools work
Review error messages - Error messages often provide clues about what's wrong
Contact support - If you're still stuck, contact MCP Functions support with details about the issue
Security Best Practices
When connecting AI assistants to your MCP servers, following security best practices protects your tools and data:
API Key Security
Never share API keys publicly - Don't post API keys in forums, code repositories, or public channels
Use different keys for different purposes - Create separate keys for development, production, and different AI assistants
Rotate keys regularly - Periodically create new keys and revoke old ones
Revoke compromised keys immediately - If you suspect a key is compromised, revoke it right away
Store keys securely - Use password managers or secure storage, not plain text files
Connection Security
Use HTTPS only - Always use the HTTPS URL provided by MCP Functions, never HTTP
Verify server URLs - Make sure you're connecting to the official MCP Functions servers
Monitor connections - Regularly check which AI assistants are connected to your servers
Review access logs - Check execution logs to see who's using your tools and when
Access Control
Limit who has API keys - Only share API keys with people who need access
Use workspace access control - Control who can see and modify MCP servers using workspace permissions
Separate development and production - Use different workspaces and servers for development and production
Regular access reviews - Periodically review who has access and remove unnecessary access
Summary: Connection Checklist
Here's a quick checklist for connecting an AI assistant to your MCP server:
✓ Create an MCP server in MCP Functions
✓ Add tools to your MCP server
✓ Get your Server URL from MCP server settings
✓ Get your API Key (copy it when first shown - you can't view it again)
✓ Open your AI assistant's settings
✓ Find the MCP/Integration section
✓ Add a new server connection
✓ Enter the Server URL and API Key
✓ Save the configuration
✓ Test by asking the AI to list available tools
✓ Test by using one of your tools
✓ Verify everything works correctly
Once connected, your AI assistant can discover and use your tools automatically. The AI will use tool descriptions to understand when to use each tool, so make sure your tool descriptions are clear and helpful.
What Happens After Connection
Once your AI assistant is connected to your MCP server:
Tools are immediately available - The AI can see and use your tools right away
Automatic discovery - When you add new tools to your server, the AI will discover them automatically (may require reconnection or refresh)
Real-time usage - The AI can call your tools whenever needed to answer user questions
Usage tracking - All tool executions are logged in MCP Functions, so you can see how your tools are being used
Your tools are now part of the AI assistant's capabilities. Users can ask questions that require your tools, and the AI will automatically use them to provide answers.
Last updated