# Connect ilert to an LLM

ilert provides[ a remote MCP server](https://github.com/iLert/mcp-ilert) that lets your AI assistant securely call ilert tools using your API key. With MCP you can unify alerting and incident-response actions inside your assistant’s workflow – faster triage, status updates, on-call escalations – which helps reduce MTTR by removing app-switching.

***

### Prerequisites

1. An ilert account and a user API key.
2. Internet access to `https://mcp.ilert.com/mcp`.
3. For Claude Desktop: latest Claude Desktop on macOS or Windows; optional Node.js if you use the `npx` wrapper approach. ([Model Context Protocol](https://modelcontextprotocol.io/quickstart/user))
4. For ChatGPT agent/deep research connectors: a plan that supports **custom connectors (MCP)** – available to Pro (with regional limits), Business, Enterprise, and Edu. See plan matrix for details. ([OpenAI Help Center](https://help.openai.com/en/articles/11487775-connectors-in-chatgpt))

***

### In ilert: create an API key

1. Log in to the **ilert app** and open your **organization settings**.
2. Go to **Profile** → **API Keys**.
3. Click **Create API Key**, copy the token, and store it securely.

***

### Add ilert mcp server directly to Cursor

Install the ilert MCP server in Cursor with one click to begin using ilert in your agent workflow.

[![Install MCP Server](https://cursor.com/deeplink/mcp-install-light.svg)](https://cursor.com/en/install-mcp?name=ilert\&config=eyJ0eXBlIjoic3RyZWFtYWJsZUh0dHAiLCJ1cmwiOiJodHRwczovL21jcC5pbGVydC5jb20vbWNwIiwiaGVhZGVycyI6eyJBdXRob3JpemF0aW9uIjoiQmVhcmVyIHtZT1VSLUlMRVJULUFQSS1LRVktSEVSRX0ifX0%3D)

### Configure your MCP client (generic)

Use **Streamable HTTP** (preferred) or the `mcp-remote` wrapper if your client doesn’t yet support Streamable HTTP.

1. Add this remote server to your MCP client configuration:

   ```json
   {
     "mcpServers": {
       "ilert": {
         "type": "streamableHttp",
         "url": "https://mcp.ilert.com/mcp",
         "headers": {
           "Authorization": "Bearer {{YOUR-API-KEY}}"
         }
       }
     }
   }
   ```

   This uses the MCP **Streamable HTTP transport** defined in the 2025-03-26 spec. ([Model Context Protocol](https://modelcontextprotocol.io/specification/2025-03-26/basic/transports?utm_source=chatgpt.com))
2. If your client does **not** support remote MCP servers or Streamable HTTP, use the stdio wrapper:

   ```json
   {
     "mcpServers": {
       "ilert": {
         "command": "npx",
         "args": [
           "-y",
           "mcp-remote",
           "https://mcp.ilert.com/mcp",
           "--header",
           "Authorization: Bearer ${ILERT_AUTH_TOKEN}"
         ],
         "env": {
           "ILERT_AUTH_TOKEN": "{{YOUR-API-KEY}}"
         }
       }
     }
   }
   ```

   This bridges stdio-only clients to a remote server via an external launcher, while passing your bearer token as a header – consistent with the MCP transport spec. ([Model Context Protocol](https://modelcontextprotocol.io/specification/2025-03-26/basic/transports?utm_source=chatgpt.com))

***

### Claude Desktop: add the ilert MCP server

#### Option A – via local config (works for stdio or with the `mcp-remote` wrapper)

1. Open **Claude Desktop** → **Settings** (from the system menu), then **Developer** → **Edit Config**. ([Model Context Protocol](https://modelcontextprotocol.io/quickstart/user))
2. Edit `claude_desktop_config.json` (auto-created) located at:
   * macOS: `~/Library/Application Support/Claude/claude_desktop_config.json`
   * Windows: `%APPDATA%\Claude\claude_desktop_config.json` ([Model Context Protocol](https://modelcontextprotocol.io/quickstart/user))
3. Insert this entry to use the wrapper:

   ```json
   {
     "mcpServers": {
       "ilert": {
         "command": "npx",
         "args": [
           "-y",
           "mcp-remote",
           "https://mcp.ilert.com/mcp",
           "--header",
           "Authorization: Bearer ${ILERT_AUTH_TOKEN}"
         ],
         "env": {
           "ILERT_AUTH_TOKEN": "{{YOUR-API-KEY}}"
         }
       }
     }
   }
   ```
4. Save the file, **quit and restart** Claude Desktop. You should see the MCP hammer indicator; clicking it shows the ilert tools. ([Model Context Protocol](https://modelcontextprotocol.io/quickstart/user))

#### Option B – as a **remote MCP server** (custom connector)

If you prefer to add ilert as a **remote** MCP server through Claude’s connector UI:

1. Open **claude.ai** → **Settings** → **Connectors** → **Add custom connector**.
2. Enter `https://mcp.ilert.com/mcp` and complete authentication using your ilert API key when prompted.
3. Manage tool permissions in the connector’s settings. ([Model Context Protocol](https://modelcontextprotocol.io/docs/tutorials/use-remote-mcp-server))

***

### ChatGPT: add ilert in agent/deep research mode

#### Add as a **custom connector (MCP)**

1. In **ChatGPT**, go to **Settings** → **Connectors** → **Add custom connector**. ([OpenAI Help Center](https://help.openai.com/en/articles/11487775-connectors-in-chatgpt))
2. Paste the ilert MCP URL: `https://mcp.ilert.com/mcp`.
3. When prompted, authenticate – for ilert, supply your **API key** as instructed by the connector flow.
4. Ensure the connector is **enabled** for your workspace or account. 

> Notes\
> • **Plan availability** and regional limits apply. Pro supports custom connectors, but some connectors and features aren’t available to users in the EEA/Switzerland/UK; Business/Enterprise/Edu have global availability.\
> • ChatGPT validates custom MCP connectors; if you see “doesn’t implement our specification”, your server likely lacks required **search** and **fetch** tools for deep research.

#### Use it in **Agent mode** / **Deep research**

1. Start a new chat and switch to **Agent mode** from the composer dropdown, or choose **Tools** → **Deep research**. ([OpenAI](https://openai.com/index/introducing-deep-research))
2. Click **Use connectors** or pick **Deep research**, then select the **ilert** connector. 
3. Ask your question or command (e.g., “Create a high-priority incident for service X and add on-call as responders”).
4. Review citations and outputs as the agent reasons over connector results. 

***

### Security and best practices

1. Scope the API key to the minimum necessary and rotate it regularly.
2. Do not hard-code tokens in shared configs; prefer environment variables.
3. Review and limit tool permissions in Claude/ChatGPT connector settings. ([Model Context Protocol](https://modelcontextprotocol.io/docs/tutorials/use-remote-mcp-server), [OpenAI Help Center](https://help.openai.com/en/articles/11487775-connectors-in-chatgpt))
4. Follow MCP transport guidance when sending headers and using Streamable HTTP. ([Model Context Protocol](https://modelcontextprotocol.io/specification/2025-03-26/basic/transports))

***

### Troubleshooting

1. **401/403 Unauthorized** – verify the `Authorization: Bearer <token>` header and that the API key is active.
2. **Claude Desktop server not visible** – restart the app; validate JSON syntax in `claude_desktop_config.json`; check logs in `~/Library/Logs/Claude` (macOS) or `%APPDATA%\Claude\logs` (Windows). ([Model Context Protocol](https://modelcontextprotocol.io/quickstart/user))
3. **Remote vs local** – if your client lacks Streamable HTTP, use the `npx mcp-remote` wrapper to bridge stdio to the ilert remote server. ([npx mcp-remote wrapper](#option-a-via-local-config-works-for-stdio-or-with-the-mcp-remote-wrapper))

***

### FAQ

**Does ilert’s MCP server support Streamable HTTP?**\
Yes – configure it with `type: "streamableHttp"` and include the bearer token in `headers`. ([Model Context Protocol](https://modelcontextprotocol.io/specification/2025-03-26/basic/transports))

**Where is the Claude Desktop config file?**\
macOS: `~/Library/Application Support/Claude/claude_desktop_config.json`\
Windows: `%APPDATA%\Claude\claude_desktop_config.json`. ([Model Context Protocol](https://modelcontextprotocol.io/quickstart/user))

**Can I add ilert as a remote connector in Claude without editing files?**\
Yes – use **Settings → Connectors → Add custom connector** in the Claude UI, then enter the MCP URL and complete auth. ([Model Context Protocol](https://modelcontextprotocol.io/docs/tutorials/use-remote-mcp-server))

**Is MCP available in ChatGPT Agent mode?**\
Yes – add ilert as a **custom connector** in ChatGPT. Use it in **Agent mode** (or via **Deep research**) by enabling connectors in Tools. Regional and plan limits apply. ([OpenAI Help Center](https://help.openai.com/en/articles/11487775-connectors-in-chatgpt), [OpenAI](https://openai.com/index/introducing-deep-research))

**Are there plan or regional restrictions in ChatGPT?**\
Custom connectors are available for Pro, Business, Enterprise, and Edu. Some connectors and features are not available to users in the EEA, Switzerland, and the UK; Business/Enterprise/Edu are available globally. ([OpenAI Help Center](https://help.openai.com/en/articles/11487775-connectors-in-chatgpt))

***