Powered byLightning AI
Documentation

    Getting started

Getting started with GraphN

This guide walks you from sign-up to a working Support Router workflow in about ten minutes. You use only built-in mock data — no external accounts or API keys for third-party services.
If you are evaluating the broader production lifecycle before building, start with how GraphN takes an AI workflow from blueprint to execution.

What you'll build

The Support Router is a pre-built blueprint that handles common customer-support questions. It includes:
  • Four agents — a triage agent that classifies the customer's intent, plus three specialists for order status, returns, and product questions
  • A mock store API with e-commerce tools (order lookup, tracking, return eligibility, product search) so every response uses real data instead of guesses
  • A routing graph that automatically hands off each conversation to the right specialist based on what the customer asks
Support Router overview in the canvas
Support Router overview in the canvas

Step 1: Sign in and create a workspace

  1. Go to graphn.ai and sign in.
  2. Create a workspace when prompted (or use the workspace switcher to add one), then make sure it's selected.
You need an active workspace before you can deploy blueprints or call APIs.

Step 2: Deploy the Support Router

  1. In the sidebar, open Blueprints.
  2. Find Support Router in the Support category (or search for it).
  3. Click the card, then choose Use this blueprint. This opens the template editor.
  4. Save the workflow (toolbar Save, or ⌘S / Ctrl+S). This creates the workflow and uploads the agents, the mock store API, and the routing graph.
  5. Publish to deploy (toolbar Publish, or ⌘⇧P / Ctrl+Shift+P). Once publish succeeds, the workflow is live in your workspace.
Blueprints library
Blueprints library
Template editor — Save and Publish
Template editor — Save and Publish

Step 3: Test your workflow

  1. Go to Home → Workflows and open Smart Customer Support That Knows Your Order History (the workflow the Support Router blueprint created).
  2. Open the Run tab.
  3. In the INPUT editor, replace the placeholder values with:
    json
    {
  "message": "Where is my order ORD-12345?",
  "order_id": "ORD-12345"
}
  4. Click Run and wait for the output to appear on the right.
The triage agent routes your question to the order specialist, which calls the mock store API to look up the order and returns real tracking details — items, total, shipping address, and carrier.
Run tab with input and output
Run tab with input and output

Step 4: Test with the API

You can call the same endpoint the app uses to run a workflow programmatically.

Get your API key

  1. Open Settings for your workspace (sidebar → Settings).
  2. Go to the API Keys tab.
  3. Create a key if you don't have one, and copy the secret when it's shown — you won't be able to see it again.

Make the request

The endpoint is:
text
POST https://cp.graphn.ai/v1/YOUR_WORKSPACE_ID/workflows/YOUR_WORKFLOW_ID/run
Replace YOUR_WORKSPACE_ID and YOUR_WORKFLOW_ID with your actual IDs (you can find both in the URL bar when viewing your workflow).
bash
curl -X POST "https://cp.graphn.ai/v1/YOUR_WORKSPACE_ID/workflows/YOUR_WORKFLOW_ID/run" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{"input":{"message":"Where is my order ORD-12345?","order_id":"ORD-12345"}}'
You can also use Python:
python
import requests

resp = requests.post(
    "https://cp.graphn.ai/v1/YOUR_WORKSPACE_ID/workflows/YOUR_WORKFLOW_ID/run",
    headers={"Authorization": "Bearer YOUR_API_KEY"},
    json={"input": {"message": "Where is my order ORD-12345?", "order_id": "ORD-12345"}},
)
print(resp.json())
The request body wraps your message inside an input object. The Authorization header uses the API key you created above.
You can also copy a ready-made snippet from the API tab on your workflow page — it fills in your workspace and workflow IDs automatically.
Workflow API tab
Workflow API tab

What's happening under the hood

  1. The triage agent reads the customer's message and classifies intent — order status, returns, or product questions.
  2. The routing graph hands off to the right specialist based on that classification.
  3. The specialist calls the mock store API tools to look up real data (orders, tracking, returns, or products).
  4. The specialist's response is returned as the workflow output.

Next steps

Get help

Stuck, found a bug, or want to talk through a workflow design? Join the GraphN Discord. The team is there, and it is the fastest way to reach us directly.
Next

Developing with agents