Timed runs

Save a search once and let your machine run it on a schedule: find new leads, draft emails, send them inside your daily cap, export the results, and send follow-ups. Nothing runs in the cloud.

#1. Give a recipe a schedule

bash

openleads recipe add miami "marketing agencies in Miami" --every "weekdays 9am" --send
openleads recipe add devs  "rust developers in Berlin"   --every "every 6h" --export csv --target devs.csv
openleads recipe add weekly "fintech founders" -n 50     --every "mon 08:30"

--every understands plain English and cron:

You write	It runs
`daily 09:00` · `9am` · `every day at 9:30pm`	once a day
`weekdays 8:30` · `weekends 10am`	Mon–Fri / Sat–Sun
`mon,wed,fri 14:00` · `tuesdays 7pm`	on those days
`hourly` · `every 2h` · `every 30m` · `every 90 minutes`	on an interval
`0 9 * * 1-5`	standard 5-field cron

openleads recipe list shows each recipe's schedule. A bad schedule is rejected when you save it, not when it should have run.

Watchers ("tell me when new leads match") take the same flag. Without one they run once a day at 09:00:

bash

openleads watch add new-clinics "dentists in Austin" --every "every 12h" --sink webhook

#2. Arm this machine once

bash

openleads schedule on        # install the agent
openleads schedule status    # is it installed, and when does each recipe run next?
openleads schedule off       # remove it

schedule on installs one small agent that wakes every 15 minutes and runs openleads drip --live. Each wake-up runs whatever is due, so one agent serves every schedule. It uses what your OS already has:

OS	Mechanism
macOS	a launchd agent in `~/Library/LaunchAgents/dev.openleads.drip.plist`
Linux	one crontab line, marked `# OpenLeads drip heartbeat`
Windows	a Task Scheduler task named `OpenLeads Drip`

Change the wake-up interval with --every-minutes 5. The old fixed daily mode (openleads schedule 09:00) still works.

If the machine is asleep or off when a run was due, the run happens at the next wake-up. It runs once, even if several slots were missed.

#3. See what happened

bash

openleads runs          # the last 20 runs
openleads runs -n 100

text

  ✓ 2026-10-07T09:00:12  recipe     miami             41.2s  found 18 · sent 12
  ✗ 2026-10-07T09:00:54  watcher    new-clinics        3.1s  URLError: network unreachable
  ✓ 2026-10-07T09:15:03  follow-ups default            0.4s  due 0 · sent 0

History lives in ~/.openleads/runs.jsonl. When a run fails, the other jobs still run and you get told about it:

a desktop notification on macOS and Linux, and
a JSON POST to notify_webhook, if you set one. Slack, Discord, ntfy and Zapier webhooks all accept it:

bash

openleads config set notify_webhook https://hooks.slack.com/services/…

#Run it in GitHub Actions instead

If you'd rather not keep a laptop awake, the OpenLeads GitHub Action runs a search on GitHub's schedule and attaches the CSV to the run:

yaml

on:
  schedule: [{ cron: "0 13 * * 1" }]
jobs:
  leads:
    runs-on: ubuntu-latest
    steps:
      - uses: Samyrrrrrr990/openleads@v4.6.1
        id: leads
        with: { query: "marketing agencies in Miami", count: 50 }
      - uses: actions/upload-artifact@v4
        with: { name: leads, path: "${{ steps.leads.outputs.file }}" }

See examples/github-action/weekly-leads.yml. Sending from CI isn't supported; the action finds leads.

#Safety

Recipes that send, and follow-ups, only go out inside the send window: weekdays, 08:00–11:00 and 13:00–16:00 local time. A sending recipe due at 7pm waits and runs at the next wake-up inside the window. Recipes that only find or export run at their exact time.
openleads drip without --live is a dry run and doesn't use up a slot.
Sends always respect the daily cap, warmup ramp, send-time windows and suppression list from Sending.
Anyone who replies or bounces is dropped from follow-ups.