We listen, so no site falls alone.
Sites in the Woods is the monitor Geeks in the Woods built to watch our own websites. It visits each one from the outside, the way a reader or an AI agent would arrive, and pings us the moment one falls over.
Other teams can use it for their sites too: sign up with an emailed code, add a site's address, and the first check runs within a minute.
Right now 7 of our sites are public here, each checked every minute. See how they're doing →
What it watches
- Is it up?
- Every homepage, every minute: whether it answers, how long the server takes to start answering, and the full breakdown from DNS to the last byte.
- Where do people and agents get stuck?
- An API recovery scan asks every API path in the site's OpenAPI file without a key (an error from one that needs a key says whether it explains how to get one) and reads the sitemap's pages (sampled where one pattern repeats hundreds of times), and says where they hit a dead end: errors that don't say how to recover, lists that don't page, pages nothing links to. It ends with fixes, for you and your coding agent, and a free re-scan confirms them. Plans from 2 sites include one scan a month (two from 10 sites, four from 25); scans need a verified domain. How the scanner behaves →
- Is it right?
- The main pages, the newest stories from each sitemap, the sitemaps themselves, page titles, and pages accidentally hidden from search. Dead ends, where a reader or an agent can't get any further: a not-found page with no way on, pages with no way back to their section, links with no label. Domain renewals and certificates, long before they run out.
- Is it fast for real readers?
- Once a day, each site loads in a real browser so we see what readers see: page weight, JavaScript errors, links that lead nowhere. Real readers' speed comes from Chrome's public data.
- Does it work for AI agents?
- Many of our sites are built for agents as much as for people, so the monitor checks the files agents rely on (llms.txt, agent cards, OpenAPI, MCP servers). On the sites made for agents, an AI agent also gets only the address and a goal, to see whether it can get there on its own. What agents need →
How it stays honest
- Outside only. No access to a site's code, logs or hosting: the monitor sees what any visitor sees. That keeps it fair, and it works for any site on any host.
- Accurate, not precise. Numbers are rounded for reading ("200 ms"), and uptime is rounded down, never up to 100%. Charts start at zero and gaps in the data show as gaps.
- No claims it hasn't earned. A site we started watching last week says "since" that day, not "90 days".
What's public
Only the sites that opt in appear here, with what a visitor would want to know: up or down, response time, uptime and past outages. The details we use to fix things stay on our team dashboard. Our own agents read the same data through a read-only API, and anyone can follow outages in the feed.
Who makes it
Geeks in the Woods is a small team in Alaska that builds startups and tools for people and places for AI agents. Sites in the Woods is one of the tools for people. We built it because we run a lot of sites, and we'd rather hear about a problem from a Discord ping at 3 a.m. than from a reader the next morning. Why we built our own →
What we learn about making sites work for agents and search engines goes into our open-source agent and search readiness audit: an open checklist, and a free readiness-audit tool anyone can run against their own site. The monitor's own checks are open source too, as agent-ready-uptime, which anyone can run for their own sites.
Questions or ideas: [email protected].
The name
If a website falls over at 3 a.m. and no one's watching, does it make a sound? Here it does. The tree in our logo stands while every site is up, falls when one is down, and fades into the fog if the monitor itself goes quiet.
Last updated .