Overview
Keragon’s MCP connection allows you to securely connect MCP-compatible AI agents to your healthcare tools so they can safely discover and use explicitly approved tools.
If your AI agent is not listed as a preconfigured client (such as ChatGPT, Claude, or Le Chat), you can use Other to create a generic MCP client connection that works with any AI agent that implements the MCP protocol.
This guide explains how to connect an AI agent using the Other option.
When to Use “Other”
Use Other when:
- Your AI agent is not ChatGPT, Claude, or Le Chat
- You are using a custom, self-hosted, or proprietary AI agent
- Your agent supports the MCP protocol
- You want full control over how the agent connects to Keragon
Note: The AI agent must implement the MCP protocol. Selecting “Other” does not add MCP support to agents that do not already support it.
Step 1: Choose “Other” as the MCP Client
- Open your MCP server in Keragon.
- Go to the Connect tab.
- In Choose client, select Other.
Selecting Other enables a generic MCP connection that can be used by any MCP-compatible AI agent.
Step 2: Copy the MCP Server URL
After selecting Other, Keragon generates an MCP server URL.
- Copy the MCP server URL shown in the Connect panel.
- Keep this URL handy. You will use it when configuring your AI agent.
The MCP server URL is the endpoint your AI agent uses to communicate with Keragon.
Step 3: Configure Your AI Agent
In your AI agent’s settings, dashboard, or codebase:
- Locate the configuration for MCP, integrations, connectors, or external tools
- Paste the MCP server URL provided by Keragon
- Ensure MCP support is enabled in the agent
⚠️ Important → Keragon is framework-agnostic, but the AI agent must correctly implement the MCP protocol, including tool discovery and invocation.
Each AI agent handles MCP setup differently, but the MCP server URL is always required.
Step 4: Configure Authentication
Keragon MCP supports configured authentication methods for MCP connections. The available options depend on:
- The MCP client’s capabilities
- How the connection is configured in Keragon
Supported methods may include:
- OAuth-based authentication
- API key authentication
- No authentication (only for trusted or internal setups)
Select the authentication option that matches your AI agent’s capabilities and complete any required setup.
⚠️ Important → Authentication availability and behavior vary by client. Always confirm which methods your AI agent supports before sharing or deploying the connection. Keragon securely stores credentials and does not expose them directly to the AI agent.
Step 5: Test the Connection
After setup:
- Your AI agent should be able to discover only the MCP tools explicitly enabled in Keragon
- Tool access is limited by the permissions you configured
- All actions performed through MCP are logged in Keragon
✅ If the agent can list or invoke approved tools, the connection is working.
⚙️ Troubleshooting
If the connection does not work as expected:
- Verify the MCP server URL is correct
- Confirm the authentication method is supported by your AI agent
- Ensure MCP is enabled and properly implemented on the client side
- Check that the required tools are enabled and permitted in Keragon
→ If you are still blocked, open a support ticket from the Connect screen in Keragon.
Security and Permissions
- Only explicitly approved tools are exposed to the AI agent
- Credentials are stored securely in Keragon, not in the client
- All MCP interactions are logged for auditability and review
Keragon ensures you remain in full control of what the AI agent can access and do.
🔐 For the complete security guide, follow the Keragon MCP & HIPAA Compliance article!
Getting Started Checklist
- Select Other as your MCP client
- Copy the MCP server URL
- Configure your MCP-compatible AI agent
- Configure authentication
- Test tool discovery and usage
- Monitor activity via Keragon logs
Summary
- What this covers: Connecting any MCP-compatible AI agent to Keragon using the Other option
- Who it’s for: Customers using custom, self-hosted, or non-preconfigured AI agents
- Outcome: A secure MCP connection with explicit permissions, credential isolation, and full visibility