Overview
What the Stacksync CLI does, how a workflow connector goes from your machine to your workspace, and where to start.
The Stacksync CLI is the command-line tool for building on Stacksync. Today it builds workflow connectors: connectors that add your own actions to Stacksync workflows, for internal systems, proprietary applications or any API that Stacksync does not cover yet.
curl -fsSL https://cli.stacksync.com/install.sh | sh
stacksync login
stacksync workflow-connector create "My App"
cd workflows-app-my-app
stacksync workflow-connector run
stacksync workflow-connector deployHow it works
Create
stacksync workflow-connector create generates a project from the official template: settings, a sample module, the guides and the rules for AI coding assistants.
Build
You write the modules in Python with the Workflow Connector Development Kit (CDK). Each module is one step a workflow can run.
Test in your workspace
stacksync workflow-connector run starts the connector on your machine and adds it to your workspace as a temporary private connector. You use it in the workflow builder like any other connector, and every request runs your local code. Saved changes reload immediately. When you stop run, the temporary connector is removed. Nothing is published.
Deploy
stacksync workflow-connector deploy sends the connector's source to Stacksync. Stacksync checks it against the connector standard, and a Stacksync reviewer approves it.
Live
Stacksync builds and hosts the connector. You get an email when it's done, and its modules appear in the workflow builder of every Stacksync workspace, not only yours. You run no servers.
The full process is described in Deploy a workflow connector.
The CLI and the CDK
| Stacksync CLI | Stacksync CDK | |
|---|---|---|
| What it is | A command-line program. | A Python library, stacksync_cdk. |
| What it does | Creates projects, runs them on your machine, deploys them. | Turns your module functions into the endpoints Stacksync calls. |
| Where it runs | Your terminal. | Inside your connector, locally and once deployed. |
| How you get it | The installer. See the Quickstart. | requirements.txt of the project. run installs it. |
Built for humans and AI agents
Every step can be done by an AI coding assistant:
- The generated project contains
AGENTS.md(andCLAUDE.md), the rules an assistant must follow, and aguide/folder that explains each part. stacksync login --api-keyandstacksync login --checkwork without prompts.python -m stacksync_cdk.connector_checkslists every problem that would block a deployment.workflow-module deleteonly works inside the connector, so an agent cannot remove a module of another project.- Each page of these docs is also available as Markdown: add
.mdto its URL.