Skip to main content
Glama

erddap-mcp

MCP server over the ERDDAP ocean-data protocol — the standard keyless data-server protocol behind NOAA CoastWatch satellite SST, Argo profiling floats, IOOS gliders, NDBC buoys, EMODnet physics sensors, NCEI archives and hundreds more.

Gap: no existing MCP server for ERDDAP (GitHub search zero hits as of Oct 2026; every "ocean data" MCP found wraps a single API — this one speaks the protocol, so it works against any ERDDAP server out of the box).

Tools

Tool

What it does

list_servers

Curated public ERDDAP servers (CoastWatch, EMODnet, IFREMER, IOOS gliders, NCEI)

search_datasets

Keyword search a server's catalogue; returns dataset IDs, titles, institutions

dataset_info

Title, summary, time/space coverage, and the full queryable variable list

query_tabledap

Tabular queries: variables + time range + lat/lon box + extra constraints (!=NaN etc.), dedup, row cap

query_griddap

Gridded queries (satellite SST/analysis fields) at a point or box; time defaults to latest grid; stride + cell-count guard

Works with any ERDDAP base URL: pass a curated nickname or any https://.../erddap URL.

Related MCP server: erddap-mcp-demo

Usage

pipx install erddap-mcp      # or: pip install erddap-mcp

# stdio (default) — for Claude Code / Cursor / any MCP client
erddap-mcp

# SSE / streamable HTTP for remote clients
erddap-mcp --transport sse --port 8000
erddap-mcp --transport http --port 8000

Claude Code config:

{"mcpServers": {"erddap": {"command": "erddap-mcp"}}}

No API keys. No configuration. Read-only.

Development

python3 -m venv .venv && .venv/bin/pip install -e '.[dev]'
.venv/bin/pytest tests/ -q

Example

> search_datasets(server="coastwatch", query="sea surface temperature")
## ERDDAP datasets matching 'sea surface temperature' (15 found)

- jplMURSST41 [Grid]  MUR SST Analysis, ...
- nceiPH53sstd1day [Grid]  NCEI Geo-Potential ...

> query_griddap(server="coastwatch", dataset_id="jplMURSST41", variable="analysed_sst",
                lat_min=36.5, lat_max=37, lon_min=-122.5, lon_max=-122)
## jplMURSST41 / analysed_sst (...)

time | latitude | longitude | analysed_sst (degree_C)
2026-10-22T09:00:00Z | 36.5 | -122.5 | 18.605
...

Notes

  • ERDDAP servers are community infrastructure: queries are cached 60 s, row counts capped (default 50, max 500) and griddap cell counts guarded, so the LLM can't accidentally pull a full raster.

  • ERDDAP's Error {code=...; message="..."} bodies are surfaced as plain readable errors (e.g. wrong variable name → the server's exact complaint).

MIT licensed.

Available Tools

5 tools
dataset_infoA

Metadata for one dataset: title, institution, summary, time/space coverage, and the full variable list you can query.

Inputs: server, dataset_id.

list_serversA

List the curated public ERDDAP ocean-data servers this MCP knows (SST, Argo floats, gliders, buoys, EMODnet…).

Inputs: no named parameters.

query_griddapB

Query a gridded dataset (satellite SST, analysis fields) at a point or lat/lon box; time defaults to the latest grid.

Inputs: time, depth, server, stride, +6 more.

query_tabledapA

Query a tabular dataset (sensors, buoys, floats, gliders): variables + time range + optional lat/lon box and extra constraints.

Inputs: server, lat_max, lat_min, lon_max, +8 more.

search_datasetsA

Search a server's dataset catalogue by keyword; returns dataset IDs with title and institution.

Inputs: limit, query, server.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.7/5.0

    Scored across 5 tools

    Disambiguation5/5

    Each tool targets a distinct stage of the ERDDAP workflow: server discovery, dataset search, metadata inspection, and two query tools that are cleanly split by dataset type (tabledap vs griddap). Descriptions explicitly state which datasets belong to which query tool, so an agent can pick correctly without guessing.

    Naming Consistency4/5

    Four of five names follow a clear verb_noun pattern (list_servers, search_datasets, query_tabledap, query_griddap) and the two query tools are symmetric. dataset_info breaks the pattern as a noun-only name, but the deviation is minor and still readable.

    Tool Count5/5

    Five tools is a tight, well-scoped surface for a data-access server: one each for discovery, search, metadata, and the two ERDDAP protocol variants. No tool is redundant and none feels missing at this scope.

    Completeness4/5

    The full core lifecycle (find server → find dataset → inspect variables/coverage → query data) is covered for both tabular and gridded data. Minor gaps remain: no export/download or output-format control, and no way to retrieve server/dataset status or error diagnostics, but agents can work around these via the query tools.

    Related MCP Connectors

    Related MCP Servers

    • F
      license
      Not graded
      quality
      D
      maintenance
      Access oceanographic and environmental data from 63+ ERDDAP servers worldwide through natural language queries. Search datasets, retrieve metadata, and download scientific data for climate research, marine biology, and coastal management.
      18
      -
    • A
      license
      A
      quality
      B
      maintenance
      Read-only access to a self-hosted GeoLens geospatial catalog: dataset search, schemas, GeoJSON features, saved maps, and sandboxed read-only SQL over PostGIS.
      9
      302
      Apache 2.0
    • A
      license
      Not graded
      quality
      C
      maintenance
      Enables keyless search of the EU Copernicus Data Space catalogue, letting you find Sentinel-1/2/3/5P, DEM, land-cover, and burnt-area products by area and date and retrieve item ids, footprints, dates, and asset paths.
      94 npm
      MIT