Plan your next pipeline
Explore your source catalog, choose the objects you need, and review a pipeline plan before creating it.
Supaflow for ChatGPT
Build and manage data pipelines in ChatGPT
Turn a pipeline request into a plan, follow a sync, or investigate a failed job. Connect your Supaflow account and work with your existing connections and pipelines through conversation.
You need a Supaflow account. Create an account or sign in.
Get started with Supaflow in ChatGPT
Start with a read-only check of your workspace. Configure credentials securely in Supaflow, then use ChatGPT to work with those connections.
Step 1
Prepare your Supaflow account
Sign in or create an account. Configure source and destination connections in Supaflow before creating or running a pipeline. Existing users can keep using their current connections.
Step 2
Install the plugin
Open the public Supaflow listing in ChatGPT and install the plugin. You can use the listing directly; no GitHub marketplace setup is needed for this path.
Step 3
Authorize Supaflow
Connect your Supaflow account when prompted and approve access in the sign-in flow. The connection uses the workspaces and actions your account permits.
Step 4
Choose your workspace and first task
Ask ChatGPT to check your identity, list your workspaces, and show available connections. Choose a workspace when needed, then inspect an existing pipeline or prepare a new one.
Your first conversation
Start with your workspace
Ask your assistant to verify your account, help you choose a workspace, and show the connections you can use. If something is missing, complete that setup in Supaflow and check again.
These are example prompts. The assistant should use your actual resources and report the results it finds.
From a pipeline idea to daily operations
Keep data moving
Run an existing pipeline, follow its job status, and manage recurring schedules.
Understand a failed run
Inspect recent jobs and reported errors, then work out the next step with your assistant.
Supaflow Code for local Codex
Need to run jobs on your own infrastructure?
When your sources need private-network access or your security requirements call for running jobs on infrastructure you control, use a private Supaflow agent. The separate Supaflow Code edition for local Codex includes CLI workflows, local MCP tools, and setup checks to help deploy and manage it. You need Node.js 18 or newer and Supaflow CLI 0.5.0 or newer. Private-agent operations also require Docker.
Supaflow Code requires a local Codex session on your machine. It is not the hosted ChatGPT plugin and does not run in browser chat or cloud sessions.
Connector development — coming soon. Guided building and testing of new connectors will extend these Code workflows. Private-agent setup is available today.
1. Install the Supaflow CLI
2. Check authentication, then sign in if needed
Before local CLI login: an organization administrator or personal-account owner must enable API access and Supaflow CLI under OAuth applications in Settings > API Keys > Permissions, review and save the settings. This local CLI-backed MCP path uses CLI permissions, independently of the hosted ChatGPT or Claude selection. If unauthenticated, run this in your own terminal:
3. Select the workspace for local operations
4. Install Supaflow Code from GitHub
If this is your first time adding the Supaflow marketplace, run:
Already added this marketplace? Refresh it instead of adding it again, then install the Code edition:
Open a fresh local Codex session after installation. Supaflow Code starts the CLI-backed MCP server as supaflow-local and uses your CLI account and selected workspace, not the hosted plugin’s OAuth connection. Do not add a second manual MCP registration with that name. If you already configured one, choose which installation to keep before connecting both.
Try a local readiness check
Private-agent deployment requires the appropriate workspace permissions. A hosted connection alone cannot perform local operations. Read the Supaflow CLI guide for prerequisites and agent setup.
Alternative: install the hosted plugin from GitHub in Codex
This installs the hosted edition in Codex. Local-agent operations still use the CLI setup above.
Connect Supaflow from the plugin experience and complete OAuth. Inspect the hosted package on GitHub.
Questions before you connect
Do I need a Supaflow account?
Yes. Create an account or sign in to your existing one, then authorize the plugin. Configure source and destination connections in Supaflow before building or running a pipeline. You can discover your workspace before those connections are ready.
When do I need a private agent?
Use a private agent when your sources need private-network access or you need to run jobs on infrastructure you control. Supaflow Code for local Codex uses the Supaflow CLI to guide deployment and troubleshooting. Install that separate edition from GitHub; the hosted ChatGPT plugin cannot operate local agents.
Can I build and test new connectors?
Guided connector development and testing in Code are coming soon. They are not included in the current hosted plugin or the setup shown here.
Does installing the plugin run a pipeline?
No. Start by inspecting your workspace and connections. Ask for a reviewed pipeline plan when you are ready to create one, and request a run separately.
How are credentials and access handled?
Authorize the hosted plugin through Supaflow sign-in. It uses your account permissions. Set up datasource credentials in Supaflow; never paste passwords, API keys, or tokens into chat. The local CLI has its own authentication and workspace selection.
What if my workspace or connections are missing?
Check the signed-in account and workspace first. If the workspace is empty, configure the source and destination in Supaflow, then ask the assistant to check again. Ask your workspace administrator for access if needed.