The daily scan

Sources

Company boards on Greenhouse, Lever, Ashby, Workday, Rippling, SmartRecruiters, Workable, BambooHR, Breezy, Personio, Teamtailor and Eightfold; the Atlassian, GitHub, Spotify and Amazon career APIs; the Google and Apple career pages (keyword-searched; they have no API); and a dozen aggregator feeds. All public, all key-free. Two of the feeds, Wellfound and Built In, read listing pages rather than an API and are off until you turn them on under Sources on the Criteria page, since both sites' terms discourage automated reading. Two optional feeds need a free key: Adzuna, which reaches listings that never make it to a company board, and USAJOBS, every federal posting in the United States.

Jobs employers post on tekjobs.timurtek.com arrive through the same scan as one more feed (openSources.tekjobs, on by default; set it to false to drop it). They are scored like every other posting; paying to post buys the listing a place in the scan, never a place in the ranking. See Posting a job.

Job alert emails saved into Inbox/ are read too, which is how LinkedIn and Indeed listings get in without anything contacting those sites.

Some sites are not reachable from a scan and the docs say so rather than pretend: Microsoft's careers API fails TLS with a certificate-name mismatch and its new site refuses unauthenticated reads; Meta's search is client-rendered. For those, paste the job page you find; the page importer reads it.

The watchlist

Targets/Companies.md is a markdown table: company, platform, slug, tier, last fetch, notes. The Sources page shows it with each board's health (failed, zero jobs, stale, never fetched, healthy), last success apart from last attempt, filters by state, a Retry failed button that fetches only those boards, and the aggregator feeds with their switches; it also lets you add a board; add_company does the same over MCP. The registry a new profile starts with is in the repository at scraper/starter/companies-table.md, and pull requests to it are the most useful contribution.

A run

npm run scan at the repository root. Around 25,000 postings fetched, deduplicated across sources on company and title, scored, and the ones that clear the bar written as notes. Listings that vanished from a board that fetched successfully are marked closed. A log note lands in Logs/, the dashboard _Home.md is regenerated, and the Runs page shows the history: each run as Success, Partial (some boards or feeds did not answer, with the list behind a disclosure) or Failed (nothing answered), whether it was the morning task, the app, an agent or the CLI, and dry runs marked; a run button, a preset picker, and Retry failed when the last run left boards behind.

Flags: --dry scores without writing, --only <slug> fetches one board, --check-slugs reports boards that answer with nothing, --criteria <name|file> scores this run with a preset.

The morning run

tekjobs schedule installs one task (Windows) or prints the one line (cron, launchd) that runs three things in a row at 07:30: the scan, the read-only mail pass, and the read-only calendar pass. The two passes need Claude Code with the Gmail and Google Calendar connectors; each leaves proposals on the Mail and Calendar pages for you to confirm, and neither sends, replies, or creates anything. A pass that cannot run (no connector, signed out) says so in data/runs.log and the next one still runs.

The journal

Every real scan ends by rewriting the last section of that day's log note, Logs/<date>.md, as a journal of the day read from the record: the runs and what they read and wrote, what was found and the best of it, every decision made that day with who made it (you, the app, an agent, the mail check, the LinkedIn export), the mail items confirmed or dismissed, and the lines added to job notes. It is rebuilt each time, so the day's journal is always the whole day so far. tekjobs journal prints it for today or any date; --write puts it in the note without a scan. The Overview page shows today's.