Ask an AI assistant whether a site is down
IsItDown.now runs a remote MCP server - the Model Context Protocol that Claude, ChatGPT, Cursor and other assistants use to call outside tools. Connect it once and your assistant can check any website from our servers, see what is down right now and look up recent outages, with a link to the evidence every time.
Connect
The server lives at one address. It needs no key and no account, and every tool only reads.
https://isitdown.now/mcp
Claude Code
claude mcp add --transport http isitdown https://isitdown.now/mcp
Claude, ChatGPT and other apps
Add a custom connector (sometimes called a remote MCP server) and paste the address above.
Cursor, Windsurf, VS Code and JSON-configured clients
{
"mcpServers": {
"isitdown": {
"type": "http",
"url": "https://isitdown.now/mcp"
}
}
}
A client that only speaks stdio can bridge to it with npx mcp-remote https://isitdown.now/mcp.
Tools
check_website - Is this website down?
Check whether any website is up right now, from IsItDown.now's servers (a different network from the user's). Answers whether the problem is the site or the user's connection, with HTTP status and timings. Needs a full address with a TLD, e.g. "example.com".
get_service_status - Status of a major service
Live status of a major online service monitored continuously (e.g. facebook.com, gmail.com, chatgpt.com, github.com): current state, whether an outage is confirmed, 30-day uptime, user reports and recent outages.
list_down_services - What is down right now
Every continuously monitored service with a confirmed outage right now, with when it started, the cause and how many users reported it.
recent_outages - Recent outages
Confirmed outages in the last N days (default 7, max 90), newest first, optionally for one service.
search_services - Find a monitored service
Search the services IsItDown.now monitors continuously by name, domain or description.
uptime_monitoring_info - Monitor my own website
How a website owner can get their own site monitored with alerts: the free plan, the Pro plan and where to sign up.
How the answers are made
The tools answer from the same measurements as this site. A major service is checked from our own network on a fixed schedule, and an outage is recorded only when two checks in a row fail - reports from the public are included, but they never declare an outage on their own. Any other website is checked live when the assistant asks, under the checker's own rules: a full address with a domain ending, public hosts only, and a per-caller rate limit.
Every result carries the status page address, so the assistant can show you the evidence rather than ask you to take its word. The methodology explains exactly what a check does and where it can be wrong.