For machines

Every page here has a file behind it. Use them freely; cite the sources you end up using.

File	What it holds
data/sources.json	All 445 source records, full fields, with verification.
`data/source/<id>.json`	One record, e.g. sodir-factpages.json.
data/datasets.json	The 10 datasets with a reader: licence, SPDX, policy, citation.
data/channels.csv	Standard channel dictionary.
data/coverage.csv	Channel coverage per dataset.
data/unit-alerts.csv	Unit errors found in source files.
data/source.schema.json	JSON Schema of a record.
llms.txt	Short guide for language models.

MCP server

Five read-only tools: search sources, list and describe datasets, list standard channels, translate a channel name.

# Claude Code, from a checkout of the package
claude mcp add open-well-data -- python -m owd.mcp_server

Command line

owd list                     # datasets with a recipe
owd get iodp-rigwatch        # download from the source, verify, standardize
owd import volve file.zip    # same, with a file you downloaded yourself
owd cite utah-forge-16b      # citation for the dataset and for Open Well Data

Recipes say where each file comes from, its size and sha256, and whether it may be downloaded automatically. Data never passes through Open Well Data: each person downloads it from the source and accepts the source's terms.

Python

import owd
owd.buscar("seismic 3d")                 # bilingual search over the catalogue
owd.canonico("Hook Load (klbs)")         # 'hookload'
owd.describir("iodp-rigwatch")            # licence, citation, how to load

The package and the MCP server are not published on PyPI yet.