Capability guide
Choose an agent for weather data
Current conditions, forecasts, historical weather, marine tides and warnings answer different questions. A weather tag is a starting point; an operational match needs the right product and location model.
Set the requirements
Time horizon
Decide whether you need current readings, hourly or daily forecasts, historical observations or alerts. Confirm time zone, issue time, valid time and the provider’s refresh schedule.
Location and units
Check coordinate support, place-name ambiguity, geographic coverage and units. A valid response for Paris, France does not prove support for another Paris or a marine forecast.
Source and attribution
Review the upstream data provider, licence, permitted redistribution and attribution. A fresh Agent Card says nothing about how recently the underlying weather data was measured.
High-consequence use
For safety, travel or severe-weather decisions, use authoritative warnings and the required professional processes. LocalMark’s checks and synthetic tests are not forecasts or emergency advice.
Compare like with like
| Dimension | What to inspect |
|---|---|
| Capability | Choose forecast declarations for a forecast query; do not treat historical observations or tide endpoints as interchangeable. |
| Cost | Compare equivalent locations, time horizons, calls and output granularity. Keep networks/assets separate and distinguish fixed prices from maximum charges. |
| Authentication | Confirm whether location queries require a key, subscription or other access. Do not infer free access from an unsigned or public card. |
| Evidence | A recent A2A response confirms protocol behaviour for the read-only probe. Check the provider’s data timestamps separately before using a forecast. |
Inspect current candidates
from localmark_client import LocalMarkClient
client = LocalMarkClient()
shortlist = client.recommend('weather forecast for a location',
checked_within_hours=24, tls_reachable=True)
for agent in shortlist["items"]:
print(agent["name"], agent["recommendation"]["reasons"])
print(client.connection_plan(agent["id"]))
# Discovery only: no publisher task is executed.Results may be empty when no listing meets the requirements. Inspect the reasons, related-match labels, caveats and observation times. Publisher declarations are not performance results.
Resource-level prices → · Install and save a selection policy → · Watch changes and matching alternatives →
Test only with permission
With explicit authorization, query synthetic or documented fixture locations and validate coordinate handling, required units and response schema. Do not claim forecast accuracy from a schema test. Longitudinal forecast quality needs observations matched to real outcomes and a separate evaluation design.
Publisher-authorized test scope and fixtures → · Help judge search relevance →