qubi-cli

CLI for qubi Agentic Flows — validate, save, and deploy workflows from the command line.

Install

pip install qubi-cli

For automated browser login (recommended):

pip install 'qubi-cli[browser]'
playwright install chromium

Quick Start

# Login (prompts for credentials once, then fully automated)
qubi login

# List workflows
qubi flow list

# Validate a workflow JSON file
qubi flow validate my_workflow.json

# Push a workflow to the platform
qubi flow save my_workflow.json --workflow-id <id>

# Download a workflow
qubi flow get <workflow-id>

# List available AI agents
qubi agents list

Commands

Authentication

On first run, qubi login prompts for your credentials and stores them at ~/.qubi/.env. After that, login is fully automated — no password prompts.

The CLI uses Playwright to navigate the full OAuth2 flow headlessly and capture the correct session cookies.

Workflow Format

Workflows are JSON graphs (React Flow format):

{
  "nodes": [
    { "id": "uuid", "type": "Start", "position": {"x": 100, "y": 200}, "data": {"type": "Start", "name": "Start"} },
    { "id": "uuid", "type": "Http", "position": {"x": 300, "y": 200}, "data": {"type": "Http", "name": "Fetch", "method": "GET", "url": "https://..."} },
    { "id": "uuid", "type": "End", "position": {"x": 500, "y": 200}, "data": {"type": "End", "name": "End"} }
  ],
  "edges": [
    { "id": "xy-edge__src-tgt", "source": "node1-id", "target": "node2-id" }
  ],
  "viewport": { "x": 0, "y": 0, "zoom": 1 },
  "executionMode": "Sequential"
}

Node Types

License

MIT