Skip to main content

Getting Started with MCP Manager

MCP Manager
Updated

This guide walks you through your first session with MCP Manager — from adding your first MCP server to seeing live traffic in your logs. By the end, you'll have a working gateway connected to your AI client with full visibility into every request.

Step 1: Add a Server

MCP servers are the tools and data sources your AI applications will access through MCP Manager. The first thing to do is add one.

If you don't have a remote server URL handy — or you're waiting on permissions to connect to an internal service — start with an open server like Context7. Context7 provides up-to-date documentation for popular libraries and frameworks, making it a practical first server for any developer. It requires no authentication, so you can get up and running immediately.

Here's how to add it:

  1. Navigate to MCP Servers in the left sidebar.
  2. Click + Add in the top-right corner.
  3. Select Remote as the server type.
  4. Enter the server URL. For Context7, use: https://mcp.context7.com/mcp
  5. For the authentication type, select Auto. MCP Manager will detect whether the server requires OAuth, token-based auth, or no authentication at all. If you're adding a server that needs specific credentials, you can choose Client Pre-Registration (for OAuth apps where you provide a client ID and secret) or Headers (for token/API key authentication) instead.
  6. If the server requires authentication, MCP Manager will prompt you to complete the auth flow — either redirecting you to the provider's authorization page or asking you to enter a token. For open servers like Context7, this step is skipped.
  7. Click Save. Your server is now registered in MCP Manager.

You should see your new server appear in the MCP Servers list with its authentication status and type.

Step 2: Create a Gateway

Servers are registered in MCP Manager, but your AI client doesn't connect to servers directly — it connects through gateways. A gateway bundles one or more servers together behind a single URL and adds logging, access control, and security policies on top.

  1. Navigate to Gateways in the left sidebar.
  2. Click + Add in the top-right corner.
  3. Give your gateway a name (e.g., "My First Gateway" or something more descriptive like "Dev Tools").
  4. Once the gateway is created, open it and go to the Assigned Servers tab.
  5. Click Assign and select the server you just added. Choose an identity scheme — for an open server like Context7, the default works fine. For authenticated servers, you'll typically want Personal so each user authenticates with their own credentials.
  6. Go back to the gateway Overview page. You'll see the Gateway URL — this is the URL your AI client will use to connect. Click the copy button next to it.

The URL will look something like:

https://mcp.mcpmanager.ai/gateway/v1/mcp?gateway=CSO-xxxxxxxx

Step 3: Connect from Your AI Client

Now paste that gateway URL into your AI client's MCP server configuration. How you do this depends on your client:

Claude Code (CLI):

claude mcp add --transport http my-gateway https://mcp.mcpmanager.ai/gateway/v1/mcp?gateway=CSO-xxxxxxxx

Claude Desktop, Cursor, or other clients with a JSON config:

Add an entry to your MCP configuration file (e.g., mcp.json or claude_desktop_config.json):

{
  "mcpServers": {
    "my-gateway": {
      "url": "https://mcp.mcpmanager.ai/gateway/v1/mcp?gateway=CSO-xxxxxxxx",
      "type": "http"
    }
  }
}

When your client starts (or restarts), it will initiate a connection to the gateway. Here's what happens next:

  1. A browser window opens — Your client kicks off dynamic client registration. You'll be taken to MCP Manager's authorization page.
  2. Authenticate with MCP Manager — Log in with your MCP Manager account. This verifies who you are and which gateways you're allowed to access.
  3. Authenticate with each server — The wizard walks you through authenticating with each MCP server assigned to the gateway. For servers using OAuth (like Atlassian or GitHub), you'll be redirected to approve access. For token-based servers, you'll enter your API key. For open servers like Context7, this step completes automatically.
  4. Confirm and redirect — After authenticating with all servers, confirm the connection. Your browser redirects back to your AI client, and the tools from your gateway are now available.

This authentication flow only happens once per connection. MCP Manager handles token refresh automatically going forward.

Step 4: Start Using Your Tools

You're connected. Try it out — ask your AI assistant something that uses the tools you just set up. If you added Context7, try asking your AI to look up documentation for a library you're working with.

Once you've made a few requests, head back to MCP Manager to see the activity:

  • Reporting — The main dashboard shows high-level metrics: total events, active connections, and usage trends over time.
  • Gateway Logs — Open your gateway and click the Logging tab to see detailed logs for every request — which tool was called, who called it, how long it took, and whether any policies were triggered.

This is the core value of MCP Manager: every interaction between your AI tools and your MCP servers is visible, attributable, and auditable.

What's Next?

You've got the basics working. Here's where to go from here:

  • Invite your team — Add users, create teams, and assign gateways so your colleagues can connect too. See the User Management article for details.
  • Add more servers — Connect to your organization's SaaS tools, deploy managed servers in your infrastructure, or set up workstation servers for local use. See the MCP Server Types article for a full walkthrough of each server type.
  • Configure policies — Set up tool provisioning to control which tools are exposed, add rules to detect and filter sensitive data, and fine-tune access controls. See the Gateways and Connections article for everything you can do with gateway policies.

Comments

0 comments

Please sign in to leave a comment.

Powered by Zendesk