Stage 1 — Job search

The Stage 1 surface lists postings the job-searcher agent surfaced for your active session date. Each card shows title, company, location bucket, score, and a star rating you set.

How results get here

Two paths feed Stage 1:

1. The job-searcher agent. Runs the sites enabled in Settings → Job Sites against your target role + work preferences. The agent isn’t a crawler — it queries each site’s search interface and reads the result pages. It writes one row per posting into jobMatches.
2. Manual add. “Add by URL” lets you paste a job-posting URL. The server tries structured extraction (JSON-LD + Open Graph) first; if those don’t yield enough fields, it falls back to the job-posting-parser agent on the page body. Either way, the row lands in jobMatches with the same shape as agent-discovered ones.

Filtering and sorting

Filter chips at the top toggle archived / favorited / recent. The sort dropdown picks between Favorites (your stars first), Relevance (the agent’s score), Location (location bucket), or Recency (posted-at when known).

Selecting a row checkmarks it. Selected rows flow into Stage 2 when you click Continue. Unselected rows stay in Stage 1 — you can come back any time via History.

When the agent missed something

If you found a posting elsewhere and want it in this session: paste the URL into “Add by URL”. The agent isn’t re-run; only your one URL is processed. The result joins the same session and goes through Stages 2-5 normally.

Searches per month

Free-tier accounts get a small monthly cap (visible in Settings → Plan). Each Stage 1 run costs one credit; the cap exists to keep the LLM bill bounded before billing wires. Manual URL adds don’t count.

Active runs and cancellation

The job-searcher run can take 1-3 minutes for a fresh search. The active-runs indicator in the header shows it’s in flight; click “Stop” if you want to cancel mid-run.