Glama – MCP Server Registry, Inspector, and Gateway

TL;DR

TL;DR: Glama is an MCP server registry with live uptime monitoring, one-click connection to 5,000+ servers, and a built-in gateway for accessing remote MCP servers directly from Claude, ChatGPT, or Cursor.

Source and Accuracy Notes

• Official site: https://glama.ai/mcp/connectors
• MCP Registry upstream: https://github.com/modelcontextprotocol/servers
• Glama is open source under Apache 2.0

What Is Glama?

Glama is a web-based MCP server registry that goes beyond a simple directory. It pulls from the official MCP Registry and adds real-time health monitoring for every listed server. Instead of installing a server locally, you can connect to it through Glama’s gateway — useful when you want to use an MCP server from a machine that cannot run it natively, or when evaluating a server before committing to a full install.

The registry covers connectors for Claude, ChatGPT, Cursor, and any other MCP-compatible client. At the time of writing, there are over 5,000 connectors indexed.

Key features:

• Live uptime monitoring — each server shows a health status derived from Glama’s periodic checks
• Gateway access — connect to any server via Glama’s relay without local installation
• Filtered search — filter by client (Claude, ChatGPT, Cursor), category, or programming language
• One-click connection params — copy the SSE endpoint URL directly into your client config

Setup Workflow

Step 1: Find a connector

Open https://glama.ai/mcp/connectors. Use the search bar or category filters to locate a server. For this walkthrough, we will use the Brave Search connector.

Step 2: Get the gateway URL

Click on the connector card. The detail page shows:

• Status — live/up/down indicator
• Endpoint — SSE URL for gateway access
• Auth — whether the server requires an API key

For Brave Search, the endpoint looks like:

https://gateway.glama.ai/brave-search/sse
```bash

### Step 3: Configure your MCP client

In Cursor, add to `~/.cursor/mcp.json`:

```json
{
  "mcpServers": {
    "brave-search": {
      "url": "https://gateway.glama.ai/brave-search/sse"
    }
  }
}
```bash

In Claude Desktop (macOS), edit `~/Library/Application Support/Claude/claude_desktop_config.json`:

```json
{
  "mcpServers": {
    "brave-search": {
      "url": "https://gateway.glama.ai/brave-search/sse"
    }
  }
}
```bash

Restart the client. The MCP server appears in the agent tool list.

### Step 4: Verify connectivity

Ask the agent to use a tool from the newly connected server. For Brave Search:

```bash
Search the web for "latest MCP protocol updates"
```bash

If the tool fires without error, the gateway connection is working.

## Deeper Analysis

### Why use a gateway instead of local install?

Local MCP servers require Node.js or Python runtime on the host machine and must be kept updated manually. Glama's gateway outsources that dependency to a hosted relay. The trade-off is latency — a gateway round-trip adds roughly 50–150ms compared to a local server. For servers you need only occasionally, this is usually acceptable.

### Server health monitoring

Glama runs periodic checks against every registered server. If a server goes down or changes its endpoint, the status indicator reflects that within minutes. This is more reliable than the official MCP Registry, which does not track server availability.

### Authentication

Some servers require an API key. Glama's gateway supports passing Authorization headers. On the connector detail page, click **Configure Auth** and enter your key — it is stored in your browser session only and never sent to Glama's servers in plaintext.

## Practical Evaluation Checklist

- Does the server respond within500ms through the gateway?
- Is the uptime indicator green on the Glama page?
- Does the server work with your specific client version?
- Are there rate limits on the free tier?

## Security Notes

- API keys entered on Glama are client-side only and never transmitted to Glama's backend
- The gateway relay acts as a proxy — the server operator can see the origin IP (Glama's relay IP), not your machine IP
- For sensitive operations, prefer a local install or a self-hosted gateway

## FAQ

**Q: Does Glama work with any MCP-compatible client?**
**A:** Yes. Any client that supports the MCP spec over SSE can connect via the gateway URL. This includes Claude Desktop, Cursor, and any custom client using the MCP SDK.

**Q: How is this different from the official MCP Registry?**
**A:** The official registry is a static list. Glama adds live health checks, gateway access, and client-specific connection instructions.

**Q: Is there a rate limit on the free gateway tier?**
**A:** Free tier allows 100 requests per day per server. Paid plans raise this to 10,000. Check the Glama pricing page for current limits.

**Q: Can I self-host the gateway?**
**A:** Glama is open source. You can deploy your own gateway relay by running the Glama server locally. See the GitHub repo for self-hosting instructions.

## Conclusion

Glama turns the MCP server ecosystem from a DIY installation problem into a managed, monitored directory with gateway access. If you are evaluating MCP servers or want to try one without a local install, the gateway is the fastest path. Bookmark [https://glama.ai/mcp/connectors](https://glama.ai/mcp/connectors) and check it before installing any MCP server from scratch.