Billing & Pricing

Agent Orchestrator uses usage-based pricing with two dimensions: tokens (tiered by model) and workflow executions.

Pricing Overview

Standard Tokens

$1

per 1M tokens

Haiku, Nova Lite, Nova Pro — fast, cost-effective models for routine tasks.

Advanced Tokens

$10

per 1M tokens

Sonnet, Nova Premier — balanced performance for most production workflows.

Premium Tokens

$50

per 1M tokens

Opus — highest capability for complex reasoning and multi-step analysis.

Workflow Executions

$0.01

per execution

Each workflow run counts as one execution, regardless of step count or model used.

How Token Usage Is Counted

Input + output tokens, summed per step

Each workflow step sends a prompt to a model and receives a response. Both the input tokens (your prompt, context, and tool results) and output tokens (the model's response) are summed together.

The total tokens for a workflow execution is the sum of all tokens across all steps. Each step's tokens are billed at the rate of the model tier used for that step.

Tiered by model

Different steps in the same workflow can use different models. Each step is billed at its model's tier rate:

5,000 tokens on Sonnet (Advanced) = 5,000 × $10/1M = $0.05
5,000 tokens on Haiku (Standard) = 5,000 × $1/1M = $0.005

How Executions Are Counted

One run = one execution

Each time you start a workflow, it counts as one billable execution, regardless of:

  • How many steps the workflow contains
  • Which models are used in each step
  • How long the execution takes
  • Whether steps run in parallel or sequentially

Model Tier Reference

Model Tier Price per 1M tokens
haiku Standard $1
nova-lite Standard $1
nova-pro Standard $1
sonnet Advanced $10
nova-premier Advanced $10
opus Premium $50

Cost Examples

Simple 2-step workflow (Haiku)

A classification workflow: extract entities, then categorize.

Step 1: 1,500 tokens × $1/1M = $0.0015
Step 2: 1,500 tokens × $1/1M = $0.0015
Execution: $0.01
Total: ~$0.013

Complex 5-step workflow (Sonnet)

A research workflow: gather, analyze, summarize, review, publish.

5 steps × ~30,000 tokens avg = 150,000 tokens
150,000 tokens × $10/1M = $1.50
Execution: $0.01
Total: ~$1.51

Mixed-tier workflow

A smart workflow: Haiku for drafts, Sonnet for final review.

Steps 1–3 (Haiku): 10,000 tokens × $1/1M = $0.01
Step 4 (Sonnet): 5,000 tokens × $10/1M = $0.05
Execution: $0.01
Total: ~$0.07

Subscription Channels

AppXen supports two ways to subscribe. Pricing and dimensions are the same for both.

Stripe (Direct)

Subscribe at console.appxen.ai/subscribe with a credit card. Usage is metered hourly and billed monthly through Stripe.

  • View usage breakdown in Settings → Billing & Usage
  • Manage payment method and invoices via Manage Billing
  • Cancel or resubscribe anytime from the console

AWS Marketplace (Coming soon)

Subscribe through the AWS Marketplace listing. Usage is metered hourly and appears on your AWS bill.

  • Consolidated billing through your AWS account
  • Manage subscription in the AWS Marketplace Console
  • Use existing AWS payment methods and purchasing workflows
  • Eligible for AWS Enterprise Discount Programs (EDPs)

Viewing Your Usage

Stripe customers

  1. Go to Settings in the console
  2. The Billing & Usage section shows your current period: tokens by tier, executions, and estimated total
  3. Click Manage Billing for invoices and payment history

AWS Marketplace customers (Coming soon)

Usage is metered automatically and appears on your AWS bill. To view details:

  1. Go to AWS Billing Console
  2. Navigate to Bills → AWS Marketplace
  3. Find "AppXen for MCP" to see metered usage by dimension

Tips for Managing Costs

Choose the right model tier

Not every step needs an advanced model. Use Haiku or Nova Lite for straightforward tasks like extraction and formatting.

Minimize tool loops

Each tool call in a step generates additional tokens. Write clear instructions so agents complete tasks in fewer iterations.

Use Haiku for drafts

Use a Standard-tier model for initial drafts, then a single Advanced-tier step for review and refinement.

Batch workflows

Process multiple items in a single workflow rather than running separate workflows for each. You pay one execution fee per run.

Frequently Asked Questions

Do failed executions count?

Yes, a failed execution still counts as one execution. However, tokens are only billed for steps that actually ran before the failure.

Are approval gates billed while waiting?

No. When a workflow pauses at an approval gate, no tokens are consumed. You're only billed for tokens when a step actively runs.

How are copilot chat tokens billed?

Copilot chat messages in the workflow editor use the same token billing. Each message exchange counts tokens at the model tier used by the copilot.

How often am I billed?

Usage is metered hourly and billed at the end of each billing period (typically monthly). Stripe customers receive a monthly invoice. AWS Marketplace customers see charges on their AWS bill.

What's Next?