aiintermediate15 min setupby Mantle

AI support agent with order lookup

A webhook hands a customer message to a tool-using agent that can look up the order over HTTP, then the answer is emailed back.

Request this template in a demo Talk to sales

Drag to pan, pinch to zoom. This is the real graph the template installs.

Purpose

A webhook hands a customer message to a tool-using agent that can look up the order over HTTP, then the answer is emailed back.

How it works

What it does

Support request — a webhook (/hooks/<tenant>/support-agent) receives {message, email, order_id?}.
Support agent — an ai.agent with an HTTP tool (get_order) reasons over the message, calling the order API only when it needs to. Every model turn and tool call is recorded in the run trace.
Email the answer — the agent's final text is emailed to the customer.

Nodes 3

01Support requesttrigger.webhookTrigger
02Support agentai.agentAI
03Email the answermail.sendDestination

View template JSON

{
  "format": "mantle-workflow/v1",
  "id": "ai-support-agent-email",
  "name": "AI support agent with order lookup",
  "graph": {
    "nodes": [
      {
        "id": "webhook",
        "type": "trigger.webhook",
        "label": "Support request",
        "config": {
          "slug": "support-agent",
          "input_schema": {
            "required": [
              "message",
              "email"
            ],
            "properties": {
              "message": {
                "type": "string"
              },
              "email": {
                "type": "string"
              },
              "order_id": {
                "type": "string"
              }
            }
          }
        },
        "worker_type": "generic-worker",
        "position": {
          "x": 60,
          "y": 220
        }
      },
      {
        "id": "agent",
        "type": "ai.agent",
        "label": "Support agent",
        "config": {
          "model": "claude-sonnet-4-5",
          "system_prompt": "You are a concise, friendly support agent for an online store. Use the get_order tool when the customer mentions an order id. Never invent order details. Answer in plain text, under 150 words.",
          "prompt": "Customer email: {{ $json.email }}\nOrder id (may be empty): {{ $json.order_id }}\n\nMessage:\n{{ $json.message }}",
          "tools": [
            "now",
            {
              "kind": "http",
              "name": "get_order",
              "description": "Fetch an order by its id.",
              "method": "GET",
              "url": "https://api.example.com/orders/{order_id}",
              "parameters": {
                "type": "object",
                "required": [
                  "order_id"
                ],
                "properties": {
                  "order_id": {
                    "type": "string"
                  }
                }
              }
            }
          ],
          "max_iterations": 6,
          "temperature": 0.2,
          "max_tokens": 800,
          "output": {
            "mode": "text"
          },
          "tool_timeout_seconds": 20
        },
        "worker_type": "ai-worker",
        "position": {
          "x": 380,
          "y": 220
        }
      },
      {
        "id": "send_reply",
        "type": "mail.send",
        "label": "Email the answer",
        "config": {
          "to": "{{ steps.webhook.email }}",
          "subject": "Re: your support request",
          "body_text": "{{ steps.agent.text }}"
        },
        "worker_type": "generic-worker",
        "position": {
          "x": 700,
          "y": 220
        }
      }
    ],
    "edges": [
      {
        "from": "webhook",
        "to": "agent"
      },
      {
        "from": "agent",
        "to": "send_reply"
      }
    ]
  }
}

Requirements

The connections and permissions each step needs, from the node contracts in the catalog.

AnthropicAPI key
Used by Support agent.
Credential: {"api_key"}
- The provider adapter is chosen from the connection; `model` must belong to it.
- Route on the AI worker (worker_type: ai-worker) so budgets, the ai_calls ledger and the circuit breaker apply.
- A tool that raises does not fail the step: the error text is returned to the model, which may retry or answer without it.
- Sub-workflow tools (kind: workflow) are reserved for a later phase.
SendGridAPI key
Used by Email the answer.
Credential: {"api_key"} or {"username", "password"}

Setup

Pick an Anthropic (or OpenAI / Google) connection on the Support agent node and set the model if you prefer another.
Point the get_order tool url at your order API; add headers for auth if needed.
Pick a SendGrid / SMTP connection on Email the answer.
Publish, then POST to the webhook URL shown on the trigger node.

Sample input and output

Synthetic examples: node-contract samples or placeholders, never customer data.

Input

The fields the template declares or reads from the request body. Values are placeholders.

{
  "message": "<message>",
  "email": "<email>",
  "order_id": "<order_id>"
}

Output

Email the answer mail.send

{
  "body": null,
  "message_id": "",
  "provider": "",
  "sent": true,
  "status_code": 0,
  "thread_id": ""
}

Failure handling

What the graph does when a step fails, read from its retry settings, error branches and error policies.

A step that fails, after any retries, fails the run. The failed step is kept with its input and error.

Solutions that use it

Customer supportCustomer support automation with OCNORA: an AI agent with email tools, order lookup, a human approval gate, per-tenant budgets and a per-iteration trace.