Playbooks · Everyday tasks · last verified 2026-10-25

Get a ranked shortlist of remote jobs in your inbox every day

For a remote worker × a daily ranked shortlist of remote jobs × n8n with the ai-daily-job-finder workflow, so a role that turns out to be US-only is caught before you apply.

Before you start

You need an n8n account or a self-hosted n8n install, an email account to send from (Gmail works, per the workflow's own notes), and an OpenRouter API key for the AI scoring. A free healthchecks.io account is optional, for an alert if the workflow stops running. The workflow file carries a placeholder key, YOUR_OPENROUTER_KEY — replace it with your own key and never commit a workflow file that holds a real key, and never share an exported workflow that does. This page was not run end to end: the workflow was read, not imported and executed, so the import steps below are the author's quick start, and the first-run checks are yours to make.

The one problem this workflow exists to solve: a posting labeled "worldwide" that turns out to be restricted once you read the fine print. The v2 workflow scores skill fit and location separately, and does not trust a "worldwide" label unless the posting text supports it. Each job is labeled eligible, restricted, or unclear, with a quote from the posting. Restricted jobs are removed from the email. That label is still a model's reading of the posting, not a guarantee — the check in step 5 is the point of the whole setup.

What "done" looks like

At the end of the first run you have one email with a ranked list of jobs, each scored from 0 to 100 with a reason, each carrying a location label and a quote. No restricted job is in the list. If the email does not arrive, or a job in it is plainly restricted, the setup is not done.

Steps

Download the v2 workflow file. From the ai-daily-job-finder repo, take the v2 file — it is the one with the five-board pull and the location check. The five boards, confirmed in the file itself this run, are Remotive, RemoteOK, We Work Remotely, Jobicy, and Himalayas. The README's v2 notes say the pull is 300+ jobs; an older breakdown in the same repo says 200+ from three boards, which describes the earlier version, not the file you are importing.
Import it into n8n. In n8n, go to Workflows, then Import from File, and select the file. Do not activate it yet — credentials and your own profile come first, or the first run scores jobs against the author's example profile.
Add your credentials. Connect your email account and put your OpenRouter key in place of the placeholder. Keep the key out of any repo, chat log, or shared export. If you use the healthchecks.io alert, connect that too; it is optional, and the workflow runs without it.
Edit the skills and keywords to match your profile. This is the step that decides whether the ranking means anything. A generic profile produces a generic ranking. Write the skills you actually have, the roles you actually want, and the locations you can actually work from — the location label is only as honest as the profile it is scored against.
Run it once by hand, then check the top three jobs yourself. Open the full posting for each of the top three jobs in the email. Read the location rules in the posting, not just the label. Does the quote in the email actually support the label? Does the posting hide a restriction further down — a tax-residency rule, a time-zone window, a "must be based in" line? If a label is wrong, that job is not a failure of your search, it is a finding about the workflow: note it, and do not apply on the strength of the label alone.
Activate it, and watch the first scheduled run. The workflow runs every few hours and emails the best matches, ranked. Check that the second email arrives on schedule, that it does not repeat jobs it already sent you — the workflow is built to skip those — and that a board being down does not silently shrink your shortlist: each board retries three times, then the run continues with the other boards, so a missing board shows up as a thinner email, not an error.
Decide the hosting question before the trial ends. n8n Cloud's Starter plan is €20 per month billed annually, with 2,500 executions a month, per n8n's pricing page read this run. The Community Edition is free and self-hosted, which means you run the server instead. The README gives no OpenRouter cost figure for this workflow, so no cost is claimed here: scoring runs one job at a time with a short pause between requests, so a run's cost grows with the number of jobs scored. Check your OpenRouter usage after the first week, not after the first day.

Check the workflow's work

Check the top three jobs by hand against their full postings before you apply to any of them. A label with a quote is evidence, not a verdict.
Check that no job in the email is restricted for your location. If one is, the location check failed on that posting, whatever the label says.
Check that the email is not repeating jobs. A repeat means the already-sent filter is not working, and the shortlist is stale.
Check that your key appears nowhere shared: not in a repo, not in an exported workflow file, not in a screenshot.

What can go wrong

A board is down and the email just gets thinner. Three retries, then the run moves on. That is resilient, but it also means a missing board looks like a quiet day. If the shortlist shrinks suddenly, check whether a board is missing before concluding the market dried up.
The "worldwide" label is wrong. That is the exact failure the v2 location check was built for, and the check can still miss a restriction buried in the posting. The hand check in step 5 is not optional polish; it is how you find out whether the check works for your locations.
The ranking reflects the wrong profile. If you skipped step 4, the scores rank jobs against someone else's skills. A high score on the wrong profile is worse than no score, because it looks authoritative.
The workflow stops and nobody notices. Without the healthchecks.io alert, a stopped workflow looks exactly like a market with no new jobs. Either connect the alert or check for the email on a schedule you will actually keep.

Limits

This workflow finds and ranks jobs. It does not apply for you, and it does not write your application — that part is still yours, and it is where the ai-job-search and career-ops pages on this site pick up.
A score from 0 to 100 is the model's judgment against your profile, not a hiring signal. A low score on a job you know well is information about the profile or the prompt, not about your chances.
The README names no specific scoring model, so no model is named here. If you change the model, the scores change, and the first week's rankings are not comparable with the second week's.

Verification

Checked 2026-10-25. The workflow was not imported or run this run; its claims rest on the author's own files, cloned depth-1 this run: the README's v2 notes (five boards, 300+ jobs, location labels with quotes, 4,000-character descriptions, three retries per board, one-at-a-time scoring with pauses) and the older breakdown (quick start, credentials, already-sent filter, healthchecks.io alert). The five boards were confirmed in both workflow files themselves. Sources: ai-daily-job-finder on GitHub and n8n's pricing page (Starter €20 per month billed annually, 2,500 executions; Community Edition free and self-hosted). Omitted: no OpenRouter cost figure is claimed, because the README gives none; no scoring model is named, because the README names none; and no claim is made that the location check catches every restriction, because that is the check you run in step 5.

Back to: Everyday playbooks hub · Also: Keep your job-hunt repo private