Remote MCP & AI Integration Guide
Step-by-step instructions for connecting ChatGPT Custom GPTs, Anthropic Claude Desktop, Cursor AI IDE, Google Gemini API, and VS Code to live Monitor Hub telemetry.
1. Overview & Architecture
The **Model Context Protocol (MCP)** is an open standard that allows AI assistants (like ChatGPT, Claude, Cursor, and Gemini) to query real-time external data safely. Monitor Hub provides a **Remote MCP Server** operating over **Server-Sent Events (SSE)** and **HTTP JSON-RPC 2.0**, secured with **OAuth 2.1 PKCE S256** and **scoped read-only access (`mcp:read`)**.
Scoped Read-Only Access
All AI tokens are strictly scoped to mcp:read. AI clients can only query monitoring data and cannot delete or modify infrastructure settings.
OAuth 2.1 PKCE Security
Implements RFC 7591 Dynamic Client Registration, PKCE S256 single-use authorization code grant, and automatic Refresh Token Rotation (RTR).
Universal Compatibility
Works seamlessly across ChatGPT Custom GPTs, Claude Desktop, Cursor AI IDE, Google Gemini, VS Code, and custom AI agents.
2. Step-by-Step Setup Guides
OpenAI ChatGPT Custom Action Setup
1. Open chatgpt.com → My GPTs → Create a GPT → Configure.
2. Under Actions, click Create new action → Import from URL → paste the OpenAPI Schema URL.
3. Select Authentication type OAuth, enter Authorization & Token URLs, Client ID: chatgpt_custom_gpt.
4. Save your Custom GPT and start querying: "Show me current server CPU load and active outages."
3. Available MCP Tools Reference (12 Read-Only Tools)
| Tool Name | Category | Description | Accepted Parameters |
|---|---|---|---|
| list_monitors | Monitors | Lists HTTP, TCP, Ping, and Keyword monitors with current status and uptime stats. | status (up|down|warning), category (uptime|api) |
| get_monitor_details | Monitors | Retrieves full diagnostic configuration, assertion settings, and last check time for a monitor. | id (integer monitor ID) |
| read_monitor_status | Monitors | Provides high-level health overview (total, up count, down count, warning count, overall uptime %). | None |
| read_server_status | Servers | Lists connected physical/virtual server agents, hostnames, OS types, IPs, and heartbeat status. | status (up|down|pending) |
| get_server_metrics | Servers | Fetches historical CPU %, RAM %, Disk %, and Network telemetry metrics over time window. | id (agent ID), hours (default: 24, max: 168) |
| read_incidents | Incidents | Retrieves active and recent outage incidents with resolution times and AI root cause analysis. | limit (default: 50, max: 100) |
| read_uptime_statistics | SLA Reports | Calculates 24-hour SLA uptime percentages and average latency statistics across monitors. | None |
| read_alerts | Alerts | Fetches recent alert log entries across Email, Slack, Discord, and Telegram channels. | limit (default: 50) |
| list_status_pages | Status Pages | Lists user public status page configurations and public URLs. | None |
| search_monitors | Search | Searches monitors by target URL, keyword, or monitor name. | query (search string) |
| list_recent_outages | Outages | Summarizes downtime events over N days with duration in minutes and root cause explanations. | days (default: 7, max: 30) |
| list_notification_channels | Alerts | Lists configured webhooks (Slack, Discord, Telegram, Email) and trigger settings. | None |
4. Troubleshooting & FAQs
Error 401: Invalid or Revoked MCP Token
Verify that your Bearer token in your config matches your API key in Dashboard Settings → AI Connectors. If using OAuth, reconnect your session.
Cors or Network Connection Timeouts
Ensure your firewall allows outgoing HTTPS requests to https://mcp.monitorhubs.com and port 443.
Connect Your AI Co-Pilot Today
Manage your AI connector tokens and active OAuth authorization sessions directly in Monitor Hub settings.
Go to AI Settings