The UptimeGlobe bot

What it is

UptimeGlobe is an uptime monitoring service. Our customers give us the addresses of websites and APIs they want watched, usually their own, and we request each one on a schedule and tell them when it stops answering properly.

We do not crawl. We never discover or index pages, and never follow links we find on a page. Each check starts from an address its owner set up on purpose. Depending on the check, it may also send:

• the redirects the site itself sends us to (owners can turn this off);
• a short, fixed sequence of requests the owner defined, such as logging in and then fetching a page, which can use values from earlier answers;
• a request to the owner's sign-in service for a token, when the address needs one;
• for AI tool servers (MCP), the few standard messages needed to confirm the server answers;
• for browser checks, which load the page in a real browser, that page's own images, scripts and styles, and any clicks or typing the owner scripted.

What we keep is the result: the status code, how long the answer took, whether any content check the owner set up passed, and the error if it failed. We do not keep copies of pages. Browser checks can save a screenshot of what the browser showed, for the customer who set up the check.

How to recognise our requests

• User agent: our HTTP checks (websites and APIs) send UptimeGlobe/1.0 (+https://uptimeglobe.com/bot). A customer can replace it on their own monitor, and browser checks identify as the browser they run in, so a request without it may still be ours.
• Signed requests (coming): we are preparing to sign our checks with HTTP Message Signatures in the Web Bot Auth format. Signed requests will carry Signature-Agent: "https://bot.uptimeglobe.com", verifiable against the public keys at https://bot.uptimeglobe.com/.well-known/http-message-signatures-directory. A signature cannot be forged without our private key, which makes it a far stronger sign than a user agent.
• No address list. Checks run from a large shared cloud network as well as from our own servers in several regions. The shared network's addresses are also used by many other services, so a list of them would not identify us.

How often we send requests

Each monitor runs on the schedule its owner chose, no more often than every 15 seconds, from up to 13 locations; some kinds of check run from one location only. When a check fails we usually repeat it once straight away to confirm the failure, so a struggling site may see two requests close together.

robots.txt

Our checks do not read robots.txt. They do not discover or index anything: each one fetches an address its owner set up deliberately, most often their own site. If you run a site and do not want it checked, tell us and we will stop (see below).

Allowing or blocking our checks

• To allow our HTTP checks through a firewall or bot filter, match the user agent above. Once signing is enabled, verifying the signature will be the reliable way.
• To block them, block the user agent. If you would rather we stopped altogether, tell us through the contact form with the address being checked. We will stop the checks and contact the customer who set them up.

Contact

Questions or complaints about our requests: our contact form. Please include the address being checked and, if you can, a timestamp from your logs.