Skip to content

Prompt recipes

Short, copy-pasteable prompt templates for common AQUAVIEW MCP scenarios. Drop these into Claude Desktop, ChatGPT, Cursor, or any MCP-enabled client.

System prompt — research assistant

You are an ocean and atmospheric data research assistant with access to
AQUAVIEW MCP, a unified STAC catalog of 700,000+ datasets across 68
sources (NOAA, IOOS, World Ocean Database, Argo, GOES-R, NEXRAD, ESA
Sentinel, IFREMER, EMODnet, Ocean Networks Canada, and more).

When the user asks about a region, prefer `bbox` over including the
region name in the free-text `q`. When the user asks "how many" or
"where are the hot spots," prefer `aggregate` over `search_datasets`.
When the user asks for download links, follow up search results with
`get_item`. Always cite the collection ID and item ID for any specific
dataset you reference.

Avoid free-text searches that aren't scoped to relevant collections —
they pull in unrelated incident reports and regulatory boundary layers.
Use `exclude_collections="INCIDENT_NEWS"` on broad searches.

User prompts

Region survey

What ocean and atmospheric data sources does AQUAVIEW have for {REGION}?
Use an aggregation to break down by collection. Then list the top 3
collections with example item IDs.

Realtime ask

What's the current {VARIABLE} near {LOCATION}? Give me the most recent
buoy or sensor reading along with the source URL.

Storm reconstruction

{STORM NAME} made landfall near {LOCATION} on {DATE}. Find me NDBC
buoys along its track that recorded extreme {VARIABLE}, GOES-R imagery
for that day, and any tropical cyclone heat potential data from AOML
for the days leading up to landfall.

Find Argo float profiles {WITHIN N KM OF / IN BBOX} {LOCATION} from
{TIME RANGE} where the float reached at least {N} dbar pressure.
Summarize temperature/salinity ranges and provide download links for
the top three.

Cross-source correlation

Compare {SOURCE_A} {VARIABLE_A} and {SOURCE_B} {VARIABLE_B} during
{EVENT} in {REGION}. Note any leads / lags between the two signals.

Catalog scoping

List the AQUAVIEW collections that contain {VARIABLE} measurements in
{REGION}. For each, give the temporal extent and an example item ID I
can fetch with get_item.

Vessel traffic / AIS

Show vessel traffic density in the approaches to {PORT NAME}. Use the
MarineCadastre AIS collection. Give me a heatmap aggregation at
geohash precision 5.

Bathymetry

What bathymetric surveys cover {REGION}? Give me the highest-resolution
DEMs available with download links.

Sea ice trend

How has sea ice extent in the {ARCTIC / ANTARCTIC} changed over the
last {N} years? Use PolarWatch and aggregate annually.

Satellite scene lookup

Find Sentinel-2 L2A scenes over {LOCATION} between {START} and {END}
with cloud cover under {N}%. Give me the COG URLs.

Tips for crafting your own

  • Always include time and place — they're the two things the agent can't infer.
  • Name the source family if you know it (e.g., "Argo / GADR", "NDBC buoys", "GOES-R imagery") to short-circuit the catalog scan.
  • Ask for download URLs explicitly when you want them — by default search results omit the asset hrefs to save tokens.
  • Use aggregate for shape questions ("how many", "where are the hot spots", "per month"). Use search_datasets for actual items.

See ../docs/prompting-guide.md for the full guide.