diff --git a/domain-skills/buildertrend/api.md b/domain-skills/buildertrend/api.md
new file mode 100644
index 00000000..a6843bf9
--- /dev/null
+++ b/domain-skills/buildertrend/api.md
@@ -0,0 +1,77 @@
+# Buildertrend — internal API map
+
+Construction PM SaaS at `buildertrend.net`. The SPA talks to two API generations,
+both same-origin cookie-authed (`credentials: 'include'`), so the fastest path is
+`fetch` from page context (`js(...)` / `page.evaluate`) after login.
+
+## Login
+
+- `https://buildertrend.net/app/...` redirects to `login.buildertrend.com/u/login` (Auth0
+  universal login) when the session is gone. Plain email + password form
+  (`input[name=username]`, `input[name=password]`), Enter submits, then it bounces back to
+  the original `/app/` URL. No MFA observed on this tenant, but treat an unexpected second
+  screen as an auth wall.
+- Sessions do expire — never assume "user is logged in" survives across days.
+
+## API generations
+
+- Legacy: `POST /api/<Entity>/Grid` (e.g. `/api/messages/Grid`) and ad-hoc GETs
+  (`/api/Messages?messageId=...`).
+- Newer: `POST /apix/v2/<Entity>/grid` (e.g. `/apix/v2/DailyLogs/grid`).
+- Both: JSON in/out, paginated via a `pagingData` object, rows newest-first, so
+  incremental scrapes can early-stop on the first already-seen row.
+
+## Daily Logs
+
+`POST /apix/v2/DailyLogs/grid`
+
+```json
+{
+  "jobIds": [44319042, ...],            // REQUIRED — empty array returns 0 rows
+  "filters": {
+    "3": "", "4": "",
+    "8": "{\"SelectedValue\":2147483647,\"StartDate\":null,\"EndDate\":null}",  // all-time
+    "10": "", "11": 0
+  },
+  "gridRequest": { "hideMultiJobsColumns": true, "emptyStateEntity": 18 },
+  "pagingData": { "pageNumber": 1, "pageSize": 20, "resetScroll": false,
+                  "firstRow": 1, "lastRow": 20, "totalRowsAllPages": 0, "currentPage": 1 }
+}
+```
+
+- Response: `{ data: [...], records: <total>, totalPages, ... }`.
+- Row fields worth knowing: `dailyLogId`, `externalId` (uuid), `jobsiteId`, `jobsiteName`,
+  `addedById`/`addedBy` (author display name), `updatedById`, `logDate`, `logTitle`,
+  `logNotes` (plain text), `tags[]`, `viewableBy {subs, owner, internalUsers}`,
+  `coordinate {latitude, longitude}`, `weatherInformation {...}`,
+  `attachedFiles.files[]` (with `thumbnail` / `docPath` URLs), `discussionLink.commentCount`.
+- The job-ids list comes from the job picker; the UI posts the full set to
+  `/api/jobpicker/SetJobPickerData` when you click "All N Open Jobs", but the grid call
+  itself carries `jobIds` explicitly — you don't need the picker call at all.
+- Date filter: key `"8"` with `SelectedValue: 2147483647` = all time. Custom-range
+  `SelectedValue`/date formats were not cracked (naive `StartDate`/`EndDate` guesses return
+  0 rows) — if you need a window, capture the request after setting the filter in the UI,
+  or just rely on newest-first + early-stop.
+
+## Jobs list / export
+
+- UI: `https://buildertrend.net/app/Jobs/List`. Grid columns include Job Name, address,
+  and Project Manager.
+- "Export all" (share icon, top right) → `POST /api/Jobsites/ExportToExcel` with
+  `{gridRequest: {selectedColumns: [...], sortColumn, ...}, filters: "<json-string>"}`;
+  the response downloads a `Jobsites.xls`. Replayable from page context for headless
+  export — capture a live payload once and reuse it.
+
+## Messages (already tooled elsewhere)
+
+- `POST /api/messages/Grid` (legacy shape, `pageSize` up to 200 works) and detail
+  `GET /api/Messages?messageId=<id>&folderId=-9`. A full incremental Playwright scraper
+  exists in the Next Modular repo at `tools/builder-trend/scrape-messages-api/`.
+
+## Traps
+
+- `jobIds: []` on DailyLogs grid silently returns `records: 0` — not an error.
+- `messageDate` on messages has no tz offset; only `messageDateForMobile` is tz-aware.
+- SPA log titles render as `<a>` without per-log hrefs — there is no scrape-able deep link
+  in the DOM; reconstruct links from ids if needed.
+- An analytics `mp-proxy/track` POST fires on most clicks — ignore it when sniffing.