Graphdom learns a task on any website from the people who do it every day, rebuilds the site offline, and turns it into automation that must pass its tests there before it touches the real thing.
Königsberg had seven bridges and an old puzzle: can you walk through the city crossing each one exactly once? Leonhard Euler proved you can't, by drawing the city as four dots and seven lines.
A website is a city of the same kind. Its pages and elements are the places, and what people do on them are the bridges. Graphdom draws that graph from your experts' own sessions.
A
B
C
D
Four steps, on a real kind of task: an import declaration on a fictional customs portal, Meridian.
Every page your experts open is a DOM: a tree of elements. Graphdom records it at every action, inside closed shadow roots and cross-origin frames too.
A recording is a walk through those trees: what was typed, what was chosen, what was clicked, and which request answered. Session 07 has 418 events, and the duties on screen are traced back to the request that carried it.
Thirty-eight sessions by five experts don't all go the same way. Where they part is where the inputs, branches and loops are: five attached a certificate of origin, three went back to fix the address.
Graphdom keeps the graph as a spider: a JSON program you can read, diff and review. It runs on Graphdom's runtime, or compiles to a Playwright script.
{ "id": "address", "do": "choose", "effect": "read", "target": "address", "query": { "input": "address" }, "cadenceMs": 90, "option": "suggestions", "commit": "click", "match": { "rule": "address-eu" }}
{ "id": "address", "do": "choose", "effect": "read", "target": "address", "query": { "input": "address" }, "cadenceMs": 90, "option": "suggestions", "commit": "click", "match": { "rule": "address-eu" }}
// compiled from spider v6, step "address"const field = page.getByRole('combobox', { name: 'Consignee address' });await field.pressSequentially(input.address, { delay: 90 });const pick = matchAddressEu(input.address);await page.getByRole('option', { name: pick }).click();await expect(field).toHaveValue(pick);
One node of seven: “choose Address”. Hover here to find it in the graph.
From the recordings, Graphdom rebuilds the portal as a sandbox: every page and every response, served with the network off. A spider is tested here, against the tests derived from your experts' sessions, before it may touch the real site.
The sandbox. Nothing leaves the machine; every answer comes from what was recorded.The real site. A spider gets here only with a passing report, and every real run becomes a new recording.
sandbox · level 2 · network off · illustrative
A spider version is accepted only with a passing test report attached. A version that submits data also waits for a person to sign it.
Spider v6 · portal.meridian-customs.example
Elena V., Operations
signed 2026-11-03
report sha256
7f3a 91c0 … c9e2
Portals change without telling you. Graphdom notices on its own, stops before anything is submitted, and its AI detects what changed and proposes the repair, unprompted. The only manual step is a person’s approval. Change the portal below and watch.
38 of 38 checks passed on the real site this morning. Now the portal ships its next release.
"targets": { "address": {+ "hosts": [{ "css": "x-address", "closed": true }], "by": [{ "label": "Consignee address" }, { "role": "combobox" }]
} }
Today a web portal is automated by a model that drives it, or by a recording that replays it. Graphdom is the third way: AI writes and repairs the automation, and never runs it.
A model reads the page and decides each click, on every run.
One recorded session, played back the same way every time.
A program learned from many sessions, proved in the sandbox before it runs.
480,000 model calls a browser agent would make. A Graphdom spider makes none.
Worked out for a 24-step task like the import declaration: runs × 24, at least one model call per step for an agent. A spider is a plain program; only a small local decision model, behind hard checks, may help it choose.
Importers, customs brokers and public bodies run on portals they don't own. Graphdom is built for the teams that answer for what those portals receive.
[[ it.d ]]
Recordings, sandboxes and the fleet run on your own infrastructure.
Node.js standard library and Chrome. No third-party runtime packages.
A spider is a plain program: same site, same inputs, same result.
Secrets are injected at run time and never written to disk.
A version that submits data waits for a person’s sign-off, report attached.
A CAPTCHA or new terms stops the run and hands it to a person.
Name the site and what the task does. We record it with your experts, build the sandbox and show you the graph and the test report before anything runs on the real site.