The technology behind every site we ship.

Point Site Forge at a business's current website and it does what a careful senior developer would do over a week, in a fraction of the time, the same way every run: probe how the site can be read, scrape every page and asset, inventory the full link graph, build the replacement, polish it for speed, prove nothing was lost, check accessibility, and deploy behind a gate.

Nine named design templates cover most local-business looks, and each build is adapted to the business's real photos, reviews, and services. The rules that matter never change: only content that actually exists, only reviews that were actually written, and never a build that loses something the old site had.

See the eight stages on the homepage.

Most website grading tools measure technical trivia. Ours measure two things a business owner cares about. COS asks how much business the current site leaves on the table: where the reviews sit, whether a phone number is reachable from every page, how the site reads on a phone, whether the design signals investment or neglect. DEO asks whether Google and AI assistants can find, understand, and recommend the business: structured data, crawlability, an llms.txt file, consistent name and address, service copy specific enough to answer a question.

The engine crawls several pages, not just the homepage, and judges what it sees with a vision model against industry profiles, so a plumber is never penalized for lacking features a plumber does not need. More than six thousand local business websites have been scored through it.

The most common way a website launch hurts a business has nothing to do with the website: someone edits DNS by hand and the inbox stops receiving mail. The go-live engine removes the hand. It reads the domain's current records, replicates them into a new zone, reads them back, and diffs the mail records against the originals. Any mismatch stops the process.

There is no AI in this path on purpose. Same input, same result, every time. The only action on the client's side is pasting two nameservers at their registrar, and if anything looks wrong, pasting the old ones back undoes it.

Third-party analytics tags are blocked by a third of browsers, slow the page, and tell you almost nothing about crawlers. Every site we host carries a small first-party beacon and an edge log instead. The beacon records how people move through the site and how fast it really was for them. The edge log records the visits people never see: search engines and AI assistants reading the site, which is the proof that the entity work is landing.

The same beacon and the same log run on this page. No ad trackers, no cross-site tracking, and the policy that says so is written to what the code actually does.

Every business has work that has to happen but does not need a person's judgment: reading what came in overnight, sorting it, pulling the right information together, writing the first version of a reply, moving a record from one system into another. That work is where mornings go. Agents do it on a schedule instead.

What keeps it safe is the boundary. The agent prepares, a person decides. Anything that reaches a customer waits for approval, and every run leaves a log showing what it touched and why.

Web standards move quickly and most advice about them is folklore. The Lab is where we test a practice on our own properties, measure it against Search Console and real-visitor data, and only then write it into the standard every client site is built and audited against. Nothing visibility-affecting changes on a client site because someone read a blog post.