All jobs

Which Ports and Terminals Changed Their Hours or Closed Today? A Freight Forwarder's Morning Advisory Check

A freight forwarder's agent reads the public advisory pages of the ports and terminals it uses each morning and lists what affects today's pickups.

Run this on RushOpens the console with this prompt ready to send.

What the agent does

This is for a freight forwarder that moves import containers from a handful of seaports to customers' warehouses by truck and rail. Every terminal posts its own notices: gate hours changed for a holiday, a gate closed for weather, an appointment system down, a vessel delayed, a rail ramp suspending a service. The dispatchers check ten or fifteen of those pages before the first trucks roll and still miss the one notice that turns a booked pickup into a dry run and a fee. This run needs no sign-in to any app: the agent uses the browser in its own cloud computer to open the public pages listed in its input, and the session returns a file.

The input lists each page with the terminal it belongs to and the forwarder's containers due there today and tomorrow, by container number, terminal and booked appointment time. The agent opens each advisory or news page, reads the notices posted since yesterday morning, and opens any notice whose title mentions gates, hours, closure, weather, appointments, vessels, rail or a holiday. A page that will not load, or that asks for a sign-in, is listed as unread with what it showed.

For each notice it records the terminal, the date and time posted, what changes, the dates and hours it covers, and a short quote of the sentence that says so, with the notice's address. It then matches notices to the forwarder's list: a container with an appointment inside a closed gate window, at a terminal whose appointment system is down, or on a vessel the notice names as delayed. Matches go first, with the container number and why.

It writes advisories.md for the dispatchers and matches.csv with one row per affected container. It does not book, cancel or move an appointment, register on any site or contact a terminal or a trucker; those calls are the dispatchers'. Where a notice is ambiguous, such as "reduced gate hours" with no times, it says so and quotes it rather than guessing the hours.

Start it with one API call

POST https://api.phoenix.so/o/{org}/sessions with this body:

{
  "agent": "claude",
  "input": "Check these terminal and port advisory pages for notices posted since 2026-11-12 06:00 America/Los_Angeles: https://www.example-port.gov/advisories (Example Port authority), https://www.example-terminal-a.com/news (Terminal A), https://www.example-terminal-b.com/notices (Terminal B), https://www.example-rail.com/intermodal/service-alerts (Example Rail ramp). For each notice about gates, hours, closures, weather, appointments, vessels, rail service or holidays, record terminal, posted time, what changes, dates and hours covered, a quote of the deciding sentence, and the notice URL. Our containers due today and tomorrow: MSKU1234565 Terminal A appointment 2026-11-13 07:30; TGHU7654321 Terminal B appointment 2026-11-13 13:00; CMAU2468104 Example Rail ramp pickup 2026-11-14 09:00. Match each container against the notices and say why it is affected. Write advisories.md (affected containers first, then every notice by terminal, then pages you could not read) and matches.csv (container, terminal, appointment, notice URL, reason). If a notice gives no hours, quote it and say the hours are not stated. Do not sign in, register, book, cancel or change anything on any site."
}
  1. Get an API key. In the console, open Settings, then API keys, and create a key for your org. Send it in the Authorization header as a Bearer token.
  2. Send the request. POST the body in request.json to https://api.phoenix.so/o/{org}/sessions, with your repository as the resources URL. The run starts in its own cloud sandbox.
  3. Read what came back. Follow the run on its live stream or wait for the run.completed webhook, then read the result and its trajectory.

What comes back

The session returns advisories.md and matches.csv as artifacts, and a short answer naming the affected containers. The dispatcher on duty opens the file before the first trucks are assigned and calls the terminal or reschedules the appointment for each match.

What the dispatcher checks: each match against the notice itself, through the link, before acting on it, since a terminal can post an update an hour later that reverses it; any page listed as unread, which they open by hand; and the containers the agent did not match, whose notices may still matter in ways the list does not show, such as a vessel delay that moves tomorrow's discharge. The file is what the pages said this morning, not a promise of what the terminal will do.

Run it on every weekday morning

The forwarder's dispatch system knows each morning which containers are due where, so its backend builds the input and sends POST /o/{org}/sessions at 5:30 a.m. local time, Monday to Friday, and reads the result from the key's run.completed webhook. A request_id that is a UUID v5 of the date keeps a retried job from starting a second run.

On a storm day the dispatcher can send the same request again at noon with the afternoon's appointments. Each run is a session of its own, so last Tuesday's check can be read later when a fee is disputed.

The use case this job follows, with its automation and trigger: Produce a report or file, no repo.

Why run it on Rush?

Terminal notices are public but scattered across fifteen sites with fifteen layouts, and the cost of missing one is a truck sent to a closed gate. The agent reads every page every morning in the same order and matches what it finds to today's containers, which a dispatcher at 5 a.m. does not have time to do. A cron job that scrapes these pages breaks the first time a terminal redesigns its site; the agent reads a redesigned page the way a person would.

Each run has its own computer and browser, so a site that sets cookies or shows a banner leaves nothing behind for tomorrow's run. The session records every page the agent opened and what its model calls cost, so when a customer disputes a dry-run fee the forwarder can show what the notice said at the time it checked.

Frequently asked questions

Some terminals post notices only inside their appointment portal. Can the agent read those?

Not in this setup: the prompt forbids signing in, and the run has no account on any portal. Those notices stay with the dispatchers, and the file lists the portal-only terminals as unread so nobody assumes they were checked.

Can it reschedule an affected appointment?

No. Rescheduling is a booking in the terminal's system with fees and slots the dispatcher has to weigh. The agent lists what is affected and why; a person decides what to move.

How do we add a terminal?

Add its public advisory page and the terminal's name to the list the backend puts in the input. Nothing else changes, because the agent reads each page as it finds it.

What if a page lists notices in a PDF?

The agent can open a linked PDF in its browser or download it to its computer and read it, then quote it the same way. A scanned image of a notice may read poorly, so the file says when the text came from a scan.

More Public websites jobs

More jobs like this: Produce a report or file, no repo

Keep reading