Open-source project
openaddresses/openaddresses avatar
openaddresses/openaddresses

OpenAddresses stores references rather than addresses, and carries three licences: CC0, BSD and whatever the source had

A global repository of open address, building, and parcel data.

3,298 stars884 forksPythonBSD-3-Clause

At a glance

What is it?
The openaddresses repository is a global register of address, parcel, footprint and street centerline data sources, not the data itself, which is downloaded elsewhere. Its source definitions are public domain, its code is BSD, and its output keeps each upstream licence. Its JavaScript manifest declares Node 10 while depending on packages that need far newer.
Who is it for?
Use this repository if you are adding a source, validating one you already have, or working out which open address data exists by country. The registry is the useful artefact and it is maintained, with the last push to the default branch on 2026-10-02 and a weekly batch run described in the developer documentation.
Can I use it commercially?
Yes. BSD-3-Clause is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
Is it still maintained?
Yes. The repository last received commits 2 days ago.
What is it written in?
Mainly Python, according to GitHub's language statistics.

Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.

From daily snapshots

Star history

  • openaddresses/openaddresses3,298 ★ · +34
Star history
Repository2026-09-242026-11-15
openaddresses/openaddresses3.3k3.3k

34 Stars in the last 52 days, measured from the daily GitHub snapshot. Compare with other projects

Editorial analysis

The repository holds references, and the data lives somewhere else

The repository description calls this a global repository of open address, building and parcel data. The brief in the file says something narrower.

It says the repository is a collection of references to address, cadastral parcel, building footprint and street centerline data sources, and it points the reader to a separate site for a data download. So what lives on GitHub is a register of where data exists and how to fetch it, not the addresses themselves.

That split runs through the whole project. There are two download surfaces named in the file: a site for a download, and a batch host holding the output of the processing pipeline. The sources directory holds one definition per source rather than the data those definitions point at.

The framing in the brief is also modest about scope. It describes the collection as open and free to use, invites people to join, download and contribute, and then says the project is just getting started. For a register that has existed long enough to accumulate a weekly batch run, that is a conservative description of its own state.

Three licences in one repository, and the metadata reports only one

The licensing section is the most careful part of the file, and it splits the project three ways.

First, the data produced by the processing pipeline is not relicensed. It says this explicitly: individual sources keep their own licences, and the team does its best to summarise those in the source definition for each source. So the output of the pipeline inherits whatever the publishing authority chose, and the registry's job is to record that rather than override it.

Second, the source definitions themselves, the JSON files in the sources directory, are released under a public domain dedication.

Third, the rest of the repository, which is the tooling, is under a BSD 3-Clause licence.

Only the third of the three appears in the repository's licence field. A tool reading that field sees BSD and concludes the data is BSD, which is exactly the conclusion the file spends a paragraph preventing.

The licence example is pinned to a commit hash, not to the branch

The one worked example in the licensing section is a link to a single source definition, and the link is more careful than most.

It points at the source JSON for one county, and it does two things. It links to the county's own open data licence, which is what a user needs in order to know what they may do with the result. And it pins the link to a specific commit hash with a line range in the fragment, rather than to the branch.

That matters because source definitions change. If the example pointed at the branch, the line range would drift the first time anyone edited the file, and the sentence in the licensing section would end up citing the wrong lines. Pinning to a commit means the illustration stays true, at the cost of pointing at a moment rather than at the current text.

It is also the only concrete illustration of the summarising practice the file describes. Everything else in the section is a policy statement about licences the registry does not itself hold.

The manifest declares Node 10 and depends on packages that need much newer

The one manifest in the repository is JavaScript, and it contradicts itself.

The package is named `oa`, its description is to manage, test and download a global register of addresses, and its engines field asks for Node 10 or newer.

Its own dependency list says otherwise. The globbing library is at version 11, and the linter is at version 9, and both of those releases require a runtime years past Node 10. There is no configuration that satisfies both the declared floor and the declared dependencies.

The floor is also the kind of value nobody sets on purpose. A constraint written as a minimum of 10 followed by a wildcard on the patch component is a leftover, not a policy, which is why it survived while every dependency moved forward.

For anyone pinning a runtime for the tooling, the dependencies are the real answer and the engines field should be ignored.

Lint and test both point at the test directory and nothing else

There are two scripts in the manifest, and both have the same shape of scope problem.

The test script runs the tape test runner over a glob covering the test directory and pipes the output through a dot reporter. The lint script runs the linter over a glob covering the same directory. Both are ES module invocations of the form the manifest's module type implies.

What neither covers is the rest of the repository. The scripts directory, the continuous integration directory, the schema directory and every source definition are outside both globs. So a source definition that fails validation, or a script that no longer runs, is not something these two commands would notice.

The validation story lives elsewhere, which is why the dependency list is what it is: a JSON schema reference parser, a validator with format support, and a glob library. Those are the pieces that let a source definition be checked against the schema directory, and the tape suite is what exercises them.

So the project's real checking story is schema validation, and the two npm scripts are the smaller half of it.

The package entry point is a file inside the test directory

Three fields in the manifest describe the repository in ways that do not match it.

The entry point is a file in the test directory. That is a library convention, and this is not a library: it is a data registry with a test suite and a schema. Whatever consumes that entry point is consuming the tests.

The version is 5.16.0, which reads like a mature release, and the repository publishes no GitHub releases at all. The number therefore exists in exactly one place, in this file, and it versions the tooling rather than the data. Nothing in the file says which schema version or which pipeline version the registry is at.

The lint configuration, by contrast, is modern: a flat config file at the repository root and version 9 of the linter with its node plugin, while the scripts still use the older glob-and-ignore style of invocation. The repository is mid-migration between two eras of tooling, which is normal and worth knowing before you file an issue about a lint failure.

Twelve organisation slots are written into the contributors section

The funding section is built from a template rather than from the file's contents.

Code contributors are acknowledged with a link to the contributors graph and a pointer to the contribution guide. Financial contributors are handled through an open collective account, with separate links for individuals and for organisations.

The organisation part is twelve links, numbered zero through eleven, each pointing at the same path with a different number in it. They are placeholders for logos and website links, listed one per line in the README so that the rendering shows the slots whether or not anything fills them.

That is why the funding section is the longest part of a short file. It also means the section carries no information about who actually funds the work, so a reader looking for the project's financial backers has to go to the collective page rather than the README.

A United States script sits at the root of a global registry

The top-level listing mixes a global registry with tooling for one country and instructions for assistants.

Beside the documentation files there is a Python script with a country in its name, applied at the root, and a directory of United States data. The rest of the layout is the machinery: a sources directory holding the definitions, a schema directory, a scripts directory, a continuous integration directory, a test directory, and configuration files for pre-commit hooks and linting.

The language split follows the same pattern. The declared primary language is Python, which fits the processing side, while the only manifest in the repository is the JavaScript one that runs the tests.

There is also a cluster of instruction files at the root: a file for agents, another for a specific assistant tool, and a directory for the first of those. Alongside them sits an outreach example document and a document describing attribute functions. A data registry that carries outreach templates and agent instructions is doing more curation than the brief suggests.

Editorial conclusion

Use this repository if you are adding a source, validating one you already have, or working out which open address data exists by country. The registry is the useful artefact and it is maintained, with the last push to the default branch on 2026-10-02 and a weekly batch run described in the developer documentation. Three things to settle before you depend on it. Which layer of licence you are actually using, because a source definition you read here is public domain, the code is BSD, and the data that comes out of the pipeline carries the upstream licence of whichever authority published it. Whether the manifest's engine floor means anything, because it declares Node 10 while the same file depends on a glob package and a linter that need far newer runtimes. And what the checks actually cover, since both the lint and the test scripts point at the test directory and nothing else, so a passing run says nothing about the scripts that produce the data.

Frequently asked questions

What is the OpenAddresses repository?

It is a global collection of references to address, cadastral parcel, building footprint and street centerline data sources, open and free to use. The repository holds source definitions rather than the data itself: downloads go through a separate site, and the output of the processing pipeline is published on a batch host. It is BSD 3-Clause licensed, not archived, publishes no GitHub releases, and the last push to master is dated 2026-10-02.

How do I add an address data source to OpenAddresses?

Open an issue giving a link to the data and a description of its coverage area, or send a pull request against the sources directory. The contribution guide at the root has the detail. The developer documentation separately covers how to test sources locally, how pull request checks work, and how the weekly batch run is structured.

What licence applies to OpenAddresses data?

Three layers. The data produced by the processing pipeline is not relicensed from the original sources, so each source keeps its own licence and the team summarises it in the source definition, with the San Francisco source cited as an example that links to the county's open data licence. The source definitions themselves are CC0 1.0, and the rest of the repository is BSD 3-Clause. Only the BSD licence of the three appears in the repository licence field.

How are OpenAddresses source definitions checked?

With a JavaScript toolchain rather than the Python pipeline: the dependencies are a JSON schema reference parser, a schema validator with format support and a glob library, and there is a schema directory at the root. The tests run with the tape runner through a dot reporter and the linter covers the same directory. Both npm scripts are limited to the test directory, so they do not cover the scripts or the source definitions.

How often is OpenAddresses data refreshed?

On a weekly batch run. The developer documentation at the root explains how the pipeline works end to end and describes the structure of that weekly run, along with how to test sources locally and how pull request checks work. No release tags are published for the registry, so the batch host is where the current output lives.

What Node version does the OpenAddresses tooling need?

The manifest declares an engine floor of Node 10 or newer, but the same file depends on a glob library at version 11 and a linter at version 9, both of which require a much newer runtime. The declared floor is a leftover constraint rather than a working minimum, so the dependencies are the reliable answer when pinning a runtime for the tests.

Official sources

  1. Issues
  2. License: BSD-3-Clause
  3. openaddresses/openaddresses on GitHub
  4. Project website
  5. README
Add this badge to your README

If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.

Add this badge to your README

markdown
[![Hysen Labs](https://hysenlabs.com/badge/openaddresses-openaddresses.svg)](https://hysenlabs.com/projects/openaddresses-openaddresses)

Star history chart for your README

markdown
[![Star history](https://hysenlabs.com/chart/openaddresses-openaddresses.svg)](https://hysenlabs.com/projects/openaddresses-openaddresses)

Was this useful?