list_collectionsLABSCO SUMMARY
The aggregation tool is what keeps a large catalogue usable.
Searching 700,000 items is only useful if you can ask about their shape before you pull any: counts per collection, a monthly histogram, a density grid at a precision you choose. That is the difference between an agent that surveys the data and one that drowns in it. The other detail worth knowing is that search requires at least one scoping argument — query text, bounding box, datetime, collections, a filter or a paging token — so an unscoped call is refused rather than returning everything.
What it is
A unified STAC catalogue served over MCP. It pulls together NOAA, all 11 IOOS Regional Associations, the World Ocean Database, the global Argo array, GOES-R, NEXRAD, ESA Sentinel, IFREMER, EMODnet, Ocean Networks Canada and more, so a question spanning several agencies is one search rather than several APIs.
What you get
list_collectionsenumerates the 68 source collections with their spatial and temporal extents, keywords and descriptions — the IDs it returns are what you scope everything else withsearch_datasetsis the workhorse: free text, a bounding box, a datetime range or an open-ended one, collection scoping and exclusion, field projection, sorting and paging by token, with a limit between 1 and 100- CQL2 filters reach per-variable column statistics, so "buoys that recorded wave heights over 6 metres" is a filter rather than a post-processing step
aggregateruns server-side counts, per-collection breakdowns, datetime histograms and geohash or geotile heatmaps — so the model never fetches raw items just to count themget_itemreturns a full record with asset download URLs pointing at NetCDF, GRIB2, GeoTIFF and Parquet files- Every tool takes an output format: csv by default for fewer tokens, toon for readability, json for programmatic consumers
Requirements
Nothing to install and no key. The server is hosted at https://mcp.aquaview.org/mcp over HTTP — one command adds it to Claude Code, and the same URL works in Claude Desktop, Cursor and any other client that takes an HTTP MCP endpoint.
Setup effort
One command — claude mcp add --transport http aquaview https://mcp.aquaview.org/mcp
USE CASES
Finding data for a specific storm
Scope to buoy and satellite collections over a bounding box and date range, and filter on recorded wave height.
Working out what exists before searching
List the collections with their coverage, then scope the search to the two or three that could hold the answer.
Getting a feel for coverage
Ask for a per-collection breakdown or a geohash heatmap instead of paging through items.
Actually downloading a file
Fetch the record and take the asset URL — NetCDF, GRIB2 or GeoTIFF — from it.
4 tools
Listed in project docs. Your client may show a different list.
search_datasetsaggregateget_itemBEFORE YOU INSTALL
What to know first
Tools you get
4 tools named in the project's documentation.
Requires
A client that can reach a remote MCP server over HTTP
INSTALLATION
Connect your client
Server setup
claude mcp add
Claude Code
claude mcp add --transport http aquaview https://mcp.aquaview.org/mcpHosted endpoint
https://mcp.aquaview.org/mcpAdd this hosted endpoint to your MCP client.
MCP Clients
Cursor
Settings → MCP → Add new MCP Server, paste the URL above, type http.