Use model properties in Power Automate, Zapier, or n8n

Use this guide if you build automations in Power Automate, Zapier, or n8n and want the properties of a Speckle model in your flow: every element’s type, name, dimensions, and parameters, as rows you can filter, count, and send on. Models uploaded to Speckle since the 2026.9 release are stored as bundles: Parquet files that low-code tools cannot read. The properties export your flow may already call (eav/query) serves only versions in the earlier format. This guide adds one small function between your flow and Speckle. Your flow sends it a project, model, and version id; it returns the version’s properties as JSON, in the same columns as the properties export.

Someone has to set the function up once: a developer, or anyone comfortable deploying a small piece of code. Where it runs depends on the tools you already use; see Choose where it runs. After that, building and changing the flow is all low-code.

How it works

Your flow’s HTTP action POSTs {"projectId", "modelId", "versionId"} to the function’s URL.
The function asks Speckle for the version’s files and reads the property files directly. Your Speckle token stays in the function’s settings, never in the flow.
Your flow gets one row per property of every element, as JSON it can parse and filter.

Each row looks like this:

{
  "object_id": "0Jgy8odqv9EeWIex0hC4$r",
  "path": "ifcType",
  "value_text": "IfcSpace",
  "value_num": null,
  "type": "string",
  "units": null,
  "internal_definition_name": null
}

Field	Holds
`object_id`	The element’s id from the source file (an IFC GlobalId, a Revit id)
`path`	The property name, such as `ifcType` or `Dimensions.Area`
`value_text`	The value as text, on every row
`value_num`	The value as a number, on numeric rows
`type`	`string`, `number`, or `boolean`
`units`	The unit, where the source gave one
`internal_definition_name`	The property’s internal definition name, when available

Type properties (Revit type parameters, IFC type properties) are included, so each element carries its type’s properties too.

Prerequisites

A Speckle personal access token with the streams:read scope, from a user who can read the project (Building with PATs)
Somewhere to run the function; see Choose where it runs
Power Automate with the premium HTTP action, Zapier with Webhooks by Zapier, or n8n

This works on Speckle Cloud and Speckle Enterprise Server 2026.9 and later. The open-source Speckle Server has no bundles: there, keep using the properties export.

Find your IDs

Open the model version in Speckle and read the IDs from the address bar:

https://app.speckle.systems/projects/{projectId}/models/{modelId}@{versionId}

If the address has no @{versionId}, the viewer is showing the latest version. Open the specific version you want so its id appears.

Choose where it runs

The function only has to be reachable over HTTPS and able to run Python, Node.js, or .NET. Pick the host your organisation already uses:

You build flows in	Simplest place for the function
Power Automate (cloud)	Azure Functions: it sits with your other Microsoft services, and the sample deploys as it is.
Power Automate Desktop	No function: a PowerShell script on the same machine writes the properties to a CSV file.
Zapier or n8n Cloud	AWS Lambda, which the sample also supports, or any serverless host you already have, such as Google Cloud Run functions (adapt the sample).
Self-hosted n8n	No separate function: run the query in a Code node, if your n8n admin enables external modules and installs `@duckdb/node-api`. The TypeScript sample’s `src/properties.ts` is the code to start from.

If your flow feeds a Power BI or Excel report rather than an automation, you may not need a function at all: Power Query can load the bundle’s Parquet files directly.

As the person who deploys the function

You do this once. The function is a ready-made sample in Python, TypeScript, and C#: examples/bundle-properties-function. Pick the language your team maintains; they behave the same.

On Azure Functions

Get just the sample folder, then publish one language folder to a Function App, for example with the Azure Functions Core Tools:

git clone --filter=blob:none --sparse --depth 1 \
  https://github.com/specklesystems/speckle-docs-new.git
cd speckle-docs-new
git sparse-checkout set examples/bundle-properties-function
cd examples/bundle-properties-function/python
func azure functionapp publish YOUR_FUNCTION_APP

For Python, create the Function App on Python 3.12. The sample’s README covers running it locally first, and the TypeScript and C# folders. Then add the token. In the Azure portal, open the Function App, go to Settings → Environment variables, and add:

Name	Value
`SPECKLE_TOKEN`	The personal access token
`SPECKLE_URL`	Only for your own server: its address

Then open the properties function, copy its function URL, and give it to the flow builder. The URL ends in ?code=..., the key that lets a flow call it, so share it as a secret.

On AWS Lambda

The Python and TypeScript folders include a Lambda handler for a function URL. Package the folder for Linux, create the function on Python 3.12 or Node.js 22, and set SPECKLE_TOKEN and a FUNCTION_KEY of your choosing as environment variables. The sample’s README has the exact package commands, handler names, and memory settings. Create a function URL with auth type NONE: the handler only answers requests that carry the FUNCTION_KEY in an x-api-key header. Give the flow builder the URL and the key, and share both as secrets.

On another host

Each sample keeps the query apart from the Azure wrapper: python/properties.py, typescript/src/properties.ts, and csharp/BundleProperties.cs run anywhere. On another host, copy that file and write the host’s own HTTP handler around it: read the three IDs from the request body, fetch the version’s file list from Speckle with the token, pass the files to the query, and return the rows as JSON. The Azure and Lambda handlers in the same folder are 30-line models to follow. Keep the token in the host’s secret settings as SPECKLE_TOKEN, and protect the URL with the host’s own key or authentication, since anyone with it can read the model’s properties.

As the flow builder

You need the function URL from whoever deployed it, and the three IDs. On AWS Lambda you also need the key, which every request sends as an x-api-key header.

Power Automate
Zapier
n8n

Add an HTTP action. Set Method to POST, URI to the function URL, and a header Content-Type: application/json. On AWS Lambda, add the header x-api-key too.
Set Body to the IDs, as shown below.
Add Parse JSON with Content set to the HTTP action’s Body. Generate the schema from the example row above, wrapped in [ ].
Add Filter array on the parsed body. For example, path equals ifcType and value_text equals IfcDoor gives one row per door. Use their object_id values in a second Filter array to pick out the rest of their properties.

Body

{ "projectId": "PROJECT_ID", "modelId": "MODEL_ID", "versionId": "VERSION_ID" }

Add a Webhooks by Zapier action with the POST event. On AWS Lambda, add x-api-key under Headers.
Set URL to the function URL and Payload Type to json.
Under Data, add projectId, modelId, and versionId.
Filter the returned rows in a Filter or Code by Zapier step.

Add an HTTP Request node with Method POST and URL the function URL. On AWS Lambda, turn on Send Headers and add x-api-key.
Turn on Send Body, set Body Content Type to JSON, and add the three IDs as fields.
Each returned row becomes an item; filter them with a Filter node.

You should see one row per property of every element in the version, as in the example above.

On Power Automate Desktop

A desktop flow runs on a Windows machine, so it needs no function: a PowerShell script on that machine fetches the version’s files, runs the query, and writes the properties to a CSV file the flow reads. Set up the machine once:

Download the DuckDB command-line tool for Windows from duckdb.org and unzip duckdb.exe to a folder such as C:\speckle.
Save Get-BundleProperties.ps1 and query.sql to the same folder.
Store the token for the Windows account that runs the flow, then restart Power Automate Desktop so it sees the new variable:
setx SPECKLE_TOKEN "YOUR_TOKEN"

Then build the flow:

Add Run DOS command with the command below, using your IDs and folder.
Add Read from CSV file on the output file, with First line contains column names turned on. Each property becomes a row of a data table.
Filter the rows with Filter data table. For example, path equals ifcType and value_text equals IfcDoor gives one row per door.

Command

powershell -NoProfile -ExecutionPolicy Bypass -File C:\speckle\Get-BundleProperties.ps1 -ProjectId PROJECT_ID -ModelId MODEL_ID -VersionId VERSION_ID -OutFile C:\speckle\properties.csv -DuckDb C:\speckle\duckdb.exe

If the script fails, Run DOS command puts the reason in CommandErrorOutput and a non-zero CommandExitCode; check them before reading the file. End -OutFile in .json instead to get the same JSON rows as the function returns.

Troubleshooting

What you see	What it means
`401` or `403`	The function’s token is missing, expired, or lacks access to that project. Check `SPECKLE_TOKEN`.
`404` from the function	The version is in the earlier format, not a bundle. For those versions, call the properties export directly, as before.
`Version has no .eav.type_eav.parquet file`	The version’s bundle is incomplete. Republish the model and try the new version.
Slow, or your flow runs out of memory	The function returns every property of every element. Narrow it: add a `WHERE` clause to the function’s query so only the rows you need come back.

GraphQL

REST

Webhooks

Guides

Use model properties in Power Automate, Zapier, or n8n

How it works

Prerequisites

Find your IDs

Choose where it runs

As the person who deploys the function

On Azure Functions

On AWS Lambda

On another host

As the flow builder

On Power Automate Desktop

Troubleshooting