# ArcGIS API for Python: Esri's documentation repository, not its library source > The Esri/arcgis-python-api repo holds notebooks, guide chapters and API reference sources. The library itself is installed from conda or ArcGIS Pro, which is the first thing to understand about it. **Esri/arcgis-python-api** — Documentation and samples for ArcGIS API for Python - Repository: https://github.com/Esri/arcgis-python-api - Website: https://developers.arcgis.com/python/ - Stars: 2,213 · Forks: 1,152 - Language: Python - License: Apache-2.0 - Published: 2026-10-07 · Updated: 2026-10-07 · Language: en - Canonical page: https://hysenlabs.com/projects/esri-arcgis-python-api ## The repository you clone is not the package you import This is the single most important thing to know before reading anything else. `import arcgis` resolves to a package installed through a package manager, not to the code in this GitHub repository. What lives here is documentation: the README states plainly that this SDK repository contains API reference documentation, samples as Jupyter notebooks, and guide chapters as Jupyter notebooks. So the workflow for most users has two halves that are easy to conflate. You install the library from Anaconda or from ArcGIS Pro, following the package managers guide on the developers site. Separately, you clone this repository when you want to read or run the samples. If you are only calling the API from your own code and never need the notebooks, cloning this repository buys you nothing at all. The repository root still carries a few files that belong to the tooling side rather than the samples side: an `environment.yml`, a `pixi.toml`, a `docker/` directory, an `update_items.py` script and an `items_metadata.yaml`. Those suggest the notebooks are built and published in a controlled way, which is consistent with what the release notes show. A sizeable share of recent commits in v2.4.3 and v2.4.4 is release-notes editing and notebook maintenance rather than library changes. ## Five ways to execute a notebook, and they are not equivalent The README lists five execution environments, and the choice between them is a real one rather than a matter of taste. Installing Anaconda and the API yourself gives you the most control. ArcGIS Pro ships its own Python environment and is the path for desktop GIS users. ArcGIS Notebooks hosted on ArcGIS Online gives you a managed environment with the credentials already handled. A Docker image exists for reproducible or headless setups. Binder will spin up a temporary environment from the repository for a quick look. The differences that matter are not the feature list, since all five can run the same notebook. They are authentication and network reach. Most ArcGIS API for Python notebooks begin by connecting to an ArcGIS Online or Portal organization, which needs a URL and credentials or an API key. The hosted notebooks route handles that for you. Everything else means you are configuring it yourself, and the authentication guide on the developers site is where that work is documented. The Docker route is also the one with a moving target. Release notes for v2.4.3 mention enabling a prerelease for the notebook Docker image, and v2.4.2 changed the base to conda-forge and PyPI. If you pin a notebook environment, pin the tag, not `latest`. ## The samples are sorted by who you are, not by topic The `samples/` directory is organized into five numbered folders, and reading them in order tells you what the library assumes about its users. `01_get_started/` is the on-ramp. `02_power_users_developers/` covers the cases where you already know ArcGIS and want the deeper surface. `03_org_administrators/` is about administering an organization rather than analysing data, which surprises people who assume the API is only about maps. `04_gis_analysts_data_scientists/` holds the analysis and deep learning notebooks. `05_content_publishers/` is about publishing and managing content items. There is also a `samples/devops_azure_functions/` directory, which points at serverless deployment, and `samples/your_first_notebook.ipynb` sitting at the top level as a single starting point. The root of the repository additionally holds one standalone notebook, `land_parcel_extraction_using_edge_detection_deep_learning_model.ipynb`, which is the sort of deep learning task the library is often asked about. What is missing from the repository is equally telling. There is no `src/` directory, no `setup.py`, no `pyproject.toml`. If you are used to reading a library's source to understand its behaviour, you will not find it here. The `apidoc/` directory is documentation sources, and the hosted reference is generated from somewhere else. ## Where ArcPy fits and why the difference matters The most common question people bring to this library is how it relates to ArcPy, and the answer is that they target different hosts. ArcPy is the module inside ArcGIS Pro and ArcGIS Desktop, tightly bound to a licensed desktop installation and its local geoprocessing tools. ArcGIS API for Python is a web client: it talks to a portal over the network, which means it can run on a laptop, in a container, on a cloud function or on ArcGIS Online with no desktop at all. That architectural difference explains the shape of the API. Operations that are local geoprocessing in ArcPy, such as running a tool on a dataset on disk, are server-side tasks in the web API, submitted as jobs and polled. The `labs/` and `misc/` directories, plus the guide chapters on 3D deep learning architectures added around v2.4.2, show the documentation side keeping pace with that model. The README also positions the library as fitting into the scientific Python ecosystem, naming Pandas, Scikit-Learn and Fast.ai as integrations, plus Jupyter. Topics on the repository confirm the intended audience: arcgis, gis, spatial-data, data-science, mapping, jupyter. If your pipeline is already a Pandas workflow, that is the pitch: spatial operations without leaving a notebook. ## What the release history says about maintenance and licensing The repository is not archived and the last push was on 2026-09-28. Three releases are visible: v2.4.4 on 2026-09-22, v2.4.3 on 2026-03-31 and v2.4.2 on 2025-10-13. That cadence is roughly every six months, which is slow for a library but normal for a documentation-heavy repository where notebooks depend on external services that change on their own schedule. The recent change lists reinforce that reading. v2.4.4 is dominated by release-notes maintenance and a notebook update for test automation. v2.4.3 is a mix of link fixes, a corrected method name, the prerelease Docker image and new living atlas guidelines. None of it looks like API surface changes. The `update_items.py` script and `items_metadata.yaml` at the root are the likely mechanism behind that housekeeping, and if you contribute samples you are entering a repository where item metadata is tracked as data. Licensing is Apache 2.0, copyright 2018 to 2026 Esri, with the full text in `LICENSE` at the root and a copy linked as `license.txt`. The README makes a point of saying the samples are yours to reuse under those terms. Keep the boundary in mind, though: the license covers this repository's contents, not the ArcGIS Online services, hosted basemaps or Living Atlas content your notebooks consume, which carry their own terms. ## Conclusion Treat Esri/arcgis-python-api as a documentation project that happens to share a repository with release engineering, not as the place where the library lives. You install the package from conda-forge or from an ArcGIS Pro environment, and you clone this repository when you want the notebooks in `samples/`, the guide chapters under `guide/`, and the API reference sources in `apidoc/`. Start at `samples/01_get_started/` if you have an ArcGIS Online organization to point at, because most of what follows assumes authentication you already have. The last push was on 2026-09-28 and v2.4.4 was released on 2026-09-22, so the notebook set is current, and the licensing is Apache 2.0 with the sample notebooks free to reuse under those terms. ## FAQ ### What are the key differences between ArcGIS API for Python and ArcPy? ArcPy runs inside a licensed ArcGIS Pro or Desktop installation and works against local data and local geoprocessing tools. ArcGIS API for Python is a web client that connects to an ArcGIS Online or Portal organization over the network, so it can run in a notebook, a container or a cloud function with no desktop present. ### Do I clone this GitHub repository to install the ArcGIS API for Python? No. The README describes this repository as documentation: API reference documentation, sample notebooks and guide chapters. The library is installed through a package manager or from ArcGIS Pro, following the install and set up guides on the developers site. ### Which Python environments can run the sample notebooks? The README lists five: local Anaconda with the API installed, ArcGIS Pro, ArcGIS Notebooks hosted on ArcGIS Online, a Dockerised environment, and Binder. They differ mainly in how authentication to your portal is configured and how much network control you keep. ### What license covers the ArcGIS API for Python samples? The repository is licensed under Apache License 2.0, copyright 2018 to 2026 Esri, and the README states that the samples can be reused under those terms. That covers this repository's contents, not the ArcGIS Online services or basemap content that many notebooks consume. ## Sources - [Esri/arcgis-python-api on GitHub](https://github.com/Esri/arcgis-python-api) - [License: Apache-2.0](https://github.com/Esri/arcgis-python-api/blob/master/LICENSE) - [Project website](https://developers.arcgis.com/python/) - [README](https://github.com/Esri/arcgis-python-api/blob/master/README.md) - [Releases](https://github.com/Esri/arcgis-python-api/releases) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/esri-arcgis-python-api