diff --git a/docs/recipes/py_arcgis_hub.md b/docs/recipes/py_arcgis_hub.md
new file mode 100644
index 00000000..9a3572d4
--- /dev/null
+++ b/docs/recipes/py_arcgis_hub.md
@@ -0,0 +1,111 @@
+# ArcGIS Hubs Harvest Recipe
+
+## Overview
+
+This article contains instructions for harvesting and ingesting metadata records from ArcGIS Hub websites. It also includes troubleshooting recommendations and information about field values.
+
+The metadata is retrieved from DCAT 1.1 API feeds using a python script, returned as a CSV using the GeoBTAA Metadata Schema, and uploaded to the BTAA Geoportal through the GeoBlacklight Admin interface.   
+
+Records can be uploaded in two ways: 
+- A *full upload*, in which all records that were found by the harvest script are published and absent records must be manually unpublished
+- A *difference upload*, in which the CSV returned by the harvest script is filtered to include only records that have changed since the previous harvest. When this *difference CSV* is uploaded, records that are newly absent in the current harvest are automatically unpublished and only newly present records are published. 
+
+This workflow uses VSCode (or another code editor) and a fastAPI browser interface to run Python scripts. This method was introduced in 2025 and replaces our Jupyter Notebook harvest recipes. 
+
+Like all of our current harvest workflows, the ArcGIS Hubs harvest script produces two output tables: a **Primary** table and a **Distributions**  table containing links. For more information, see documentation on **Distributions** (in progress). 
+
+## Workflow
+> **Important:** The following instructions assume you've already completed the **Harvest Tools Setup Guide** (in progress).  
+
+Updated January 2026.
+
+### Step 1: Launch the fastAPI ArcGIS Metadata Harvester
+1. In VSCode (or the code editor of your choice), open a bash terminal.
+2. Activate your fastAPI harvester conda environment and change directories to the **harvester-api** repository.
+3. Enter the command: **uvicorn main:app --reload**
+4. In a web browser, visit **localhost:8000**.
+5. Click the **ArcGIS Harvester** link. 
+### Step 2: Retrieve and Prepare the Hubs List
+> The list of currently active ArcGIS Hubs may change in between harvests, so it's best practice to download a new list every time. 
+1. Click the link labeled **Filter ArcGIS Hubs in GBL Admin** to open a list of all currently active ArcGIS Hubs in GeoBlacklight Admin in a new tab. 
+2. Select all results that match this search, then export and download their Primary CSV.
+3. On the fastAPI ArcGIS Harvester page, click the **Choose File** button under **Step 1: Upload CSV of ArcGIS Hubs**, navigate to and select the CSV downloaded in the previous step, and click **Upload**.
+>This interface will automatically move the CSV into the **inputs** folder in the **harvester-api** repository and rename it **arcHubs.csv**.  
+>Alternatively, any of the above steps can be done manually: 
+1. In GeoBlacklight Admin, apply the following filters: Publication State = **published**, Resource Class = **Websites**, Harvest Workflow = **py_arcgis_hub**. For added validation, you may also filter for Accrual Periodicity = **Weekly** and/or Source Platform = **ArcGIS Hub**. 
+2. Select all results that match this search, then export and download their Primary CSV.
+3. Rename the resulting CSV "**arcHubs.csv**" and move it into the **inputs** folder within the **harvester-api** respository.
+
+### Step 3: Run the Harvester
+1. Click the **Run Harvester** button. A line for each hub site will be displayed as it is harvested. A full harvest often takes about 10-15 minutes. 
+2. When the harvester has finished, confirm that both **Primary** and **Distributions** CSVs for the present date exist within the **outputs** folder in the repository. These are the *full upload* CSVs.
+
+### Step 4: Prepare and Upload the Harvest CSVs
+> **Important:** There are two ways to complete this step.
+> - A *difference upload* should done for most weekly harvests. This updates only child records that have been added or removed since the last harvest. 
+> - A *full upload* should be done once per quarter or if you don't have CSVs from a previous harvest. This updates all child records. 
+
+#### Weekly: Difference Upload
+
+How it works:
+- Compares the current full list of harvested records to the CSVs from the previous harvest saved on your computer. 
+- Returns a CSV that will unpublish newly absent records and publish newly present records. 
+- Does not update the "date accessioned" field. 
+
+Steps: 
+1. In VSCode, navigate to the **scripts** folder within the **harvester-api** repository.
+2. Open and run the **build_uploads.py** script. When it's finished, there should be new Primary and Distributions CSVs in the **outputs** folder named with the present date and the suffix "**_upload**." These are *difference* CSVs, and they include only records that have changed 
+3. In the VSCode terminal, the script should also report how many records were added and retired. Record these numbers in the harvest ticket. 
+4. In GeoBlacklight Admin, navigate to the **Admin Tools** menu and select **Import Primary**. Name the import "**[Ticket #]-ArcHubs Harvest-YYYY-MM-DD**". Select **BTAA CSV** under **Type**. 
+5. Click the **Choose file** button and select the current **primary_upload** difference CSV from the **outputs** folder. Click through to start the import.
+6. Next, navigate to the **Admin Tools** menu and select **Import Distributions**. Name the import "**[Ticket #]-ArcHubs Harvest-YYYY-MM-DD-Distributions**". *Note*: you must import the Primary CSV before the Distributions CSV. 
+
+#### Quarterly: Full Upload
+
+How it works:
+- All records that were found by the harvest script are published.
+- Records that were previously published but are now absent must be manually unpublished.
+- A previous harvest's CSV is not required, and there is no need to run the build_uploads script. 
+- Full uploads will take much longer to import in GBL admin than difference uploads.   
+
+Steps:  
+
+1. In GeoBlacklight Admin, navigate to the **Admin Tools** menu and select **Import Primary**. Name the import **[Ticket #]-ArcHubs Harvest-YYYY-MM-DD**. Select **BTAA CSV** under **Type**. 
+2. Click the **Choose file** button and select the current date's **primary** CSV from the **outputs** folder. (No need to run the build_uploads script - just use the harvest script's output directly!) Click through to start the import.
+3. Next, navigate to the **Admin Tools** menu and select **Import Distributions**. Name the import "**[Ticket #]-ArcHubs Harvest-YYYY-MM-DD-Distributions**". *Note*: you must import the Primary CSV before the Distributions CSV.
+4. Click the **Choose file** button and select the current date's **distributions** CSV from the **outputs** folder (again, NOT distributions-uploads, just [date]-distributions). Click through to start the import. 
+5. Unlike a difference upload, a full upload does not unpublish records that have been removed from their parent hubs. This must be done manually. (Details in progress.)
+
+
+## Troubleshooting
+
+If an individual hub produces an error message during a harvest, it's often worth simply ignoring it once and trying again the following week. If it continues to return an error, check if its endpoint URL still loads a valid JSON. If not, unpublish the site. Create a GitHub issue, put it on hold, and attach the "monitor" label to it.  
+
+If the Hubs list needs additions or changes, visit the article **Adding and Updating ArcGIS Hubs** (in progress). 
+
+## Explanation of Fields and Parameters
+
+Most of the following fields apply to *the child records harvested from the Hub sites*, not the Hub sites themselves. For a detailed explanation of Hub website metadata, visit **Adding and Updating ArcGIS Hubs**(in progress). 
+
+Unless otherwise specified, all values in the fields below are automatically populated by the harvest script using a combination of the parent Hub's metadata and the metadata retrieved from the API feed. 
+
+### Provenance Fields
+
+- Accrual Method: Automated retrieval
+- Accrual Periodicity: Weekly
+- Date Accessioned: The most recent date on which a record's *parent hub* was harvested. *Not necessarily* the most recent time the child record was confirmed to still exist during a harvest!
+- Date Retired: The most recent date on which a record was unpublished. 
+- Endpoint Description: DCAT API
+- Endpoint URL: The DCAT 1.1 API feed URL of the parent Hub, which can be found by clicking the "Explore Feeds" link at the bottom of a Hub's homepage. Typically, this is the hub's main URL followed by "api/feed/dcat-us/1.1.json"
+- Harvest Workflow: py_arcgis_hubs
+- Source Platform: ArcGIS Hub
+
+> Notes about the Date Accessioned field: Because records are only included in the difference CSV (described in Step 3) if they are being added or removed, the Date Accessioned won't reflect the last time a resource was found during a difference harvest - that date will be reflected by the Hub site's Date Accessioned. In other words, records that are still published were validated by the last harvest process, although they weren't changed. *When a full harvest upload is completed, the Date Accessioned field on all child records will be updated.*
+
+### Other Fields
+- Resource Class: Web Services
+- Format: The type of web service. 
+
+### Other Parameters
+- Download links are no longer included in the harvest. 
+- Distributions URLs correspond to the web services only. These are typically REST service URL. 
\ No newline at end of file