# Find the ngrok MCP server in Claude's connector directory

> The ngrok MCP server is listed in Claude's connector directory, so you can connect it to Claude on the web, Claude Desktop, and Cowork in a few clicks instead of pasting its URL into a custom connector.

- Date: 2026-11-17
- Shipped by: Alexander Kinn
- Products: Gateway
- URL: https://ngrok.com/changelog/mcp-claude-directory

The [ngrok MCP server](https://ngrok.com/blog/mcp) is now listed in [Claude's connector directory](https://claude.ai/directory/ngrok-mcp). Search for **ngrok**, click **Connect to Claude**, sign in, and Claude can read your endpoints, Traffic Inspector traffic, and usage. Within the scopes you approve, it can also update your Traffic Policy.

![The ngrok listing in Claude's connector directory, labeled New and Community. It describes the connector as a way to inspect and manage ngrok endpoints, traffic, and deployments from Claude, lists tools such as docs_search, gw_list_endpoints, and gw_create_ip_policy, and shows the connector URL https://mcp.ngrok.com/mcp with sign-in required.](https://ngrok.com/blog-assets/images/2026-11-17-mcp-claude-directory/ngrok-claude-directory-listing.png)

Until now, connecting Claude on the web, Claude Desktop, or Cowork meant adding a custom connector by hand and typing `https://mcp.ngrok.com/mcp` exactly, `/mcp` and all. The directory listing fills in the URL for you. It also makes ngrok eligible for Claude's in-chat connector suggestions, so Claude can offer it when you ask about your tunnels and haven't connected it yet.

The connector is the same MCP server your coding agent uses. It works on every ngrok plan, including free, and it can't do more than your role in your ngrok account allows. You sign in with OAuth and approve the scopes on a consent screen, so you never paste an API key or authtoken into Claude.

Claude labels the listing **Community**. That means Anthropic ran automated checks on it but doesn't verify how its tools work. The ngrok MCP server [marks each tool](https://ngrok.com/docs/mcp-server#make-changes-safely) as read-only or as one that makes changes, and Claude lets you set each tool to **Always allow**, **Needs approval**, or **Blocked**. Keep approval on for anything that changes your account. The [tools and scopes reference](https://ngrok.com/docs/mcp-server/tools) lists every tool, whether it makes changes, and the scope it needs.

On Claude Team and Enterprise plans, an organization Owner adds the connector for the organization first. Each member then connects with their own ngrok account.

## Get started

You need a Claude account and an ngrok account. [Sign up for ngrok](https://dashboard.ngrok.com/signup) if you don't have one yet.

<Steps>
<Step title="Open the directory">

In Claude, go to **Customize > Connectors** and select **Discover**. Or open the [ngrok listing](https://claude.ai/directory/ngrok-mcp) directly.

</Step>
<Step title="Connect ngrok">

Search for **ngrok**, open the listing, and click **Connect to Claude**. On a Team or Enterprise plan, an Owner sees **Connect for your team** instead, and members see **Request** until an Owner adds it.

</Step>
<Step title="Sign in and approve">

Sign in to ngrok and approve the scopes on the consent screen. Back in Claude, ngrok shows up under **Your connectors** as **Connected**.

</Step>
<Step title="Ask it something">

Start a chat and ask "Which ngrok account am I signed in to?" Claude replies with your user, account, role, and plan. Then try something real, like "Which of my endpoints got the most traffic yesterday?" The [example prompts](https://ngrok.com/docs/mcp-server/prompts) have more ideas.

</Step>
</Steps>

Using Claude Code, Codex, or Cursor instead? The same server works there too. [Pick your client](https://ngrok.com/mcp) for its setup command.
