for ai & llm workflows
Ask an AI what’s broken, over MCP
Point a Model Context Protocol client (Claude, an IDE, anything that speaks MCP) at your monitoring and ask it what’s down in plain language. Read tools answer from your real monitors; write tools take action only behind your explicit approval. Same tenant isolation, scopes and rate limits as the dashboard.
what you get
- MCP endpoint mcp.uptimepage.dev/mcp
- Connect OAuth one-click, or scoped token
- Tools 13 (read + fenced writes)
- Every write your approval + an audit row
- Clients Claude, IDEs, any MCP client
- Price to start free, no card
ask your monitoring in plain language
Read tools hand the model the same forensics a good engineer reaches for: which monitor is down and since when, an incident’s full timeline, and why a check is slow: DNS, connect, TLS handshake and time-to-first-byte reported separately, so "slow because TLS" and "slow because DNS" come back as different answers.
actions stay behind a human
Most tools can only look. The few that act (run a check, pause or resume a monitor, post to an incident) can’t fire without a scoped token, your in-the-moment approval naming the exact effect, and an audit row for every outcome. There is no "remember my choice."
your data can’t hijack the assistant
A monitor name or scraped error text is written by someone else, and now an LLM is reading it. Every piece of customer-supplied text reaches the model labelled as data to report, never instructions to act on. Even a fooled model still can’t act without your out-of-band approval.
one-click OAuth, no copy-paste
Your client discovers the server, you log in with the session you already have, approve a consent screen, and a scoped, org-bound, expiring token is minted behind the scenes. The one lifetime the consent screen won’t offer is "never expires."
Point an MCP client at the server
{
"mcpServers": {
"uptimepage": {
"url": "https://mcp.uptimepage.dev/mcp"
}
}
}