Skip to content

copilot-agents-cost-tool

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

copilot-agents-cost-tool

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
GUIDE.md
GUIDE.md
 
 
README.md
README.md
 
 
index.html
index.html
 
 

README.md

M365 Copilot Agents Cost Calculator

A self-contained, browser-based cost estimator for Microsoft 365 Copilot agents. Open AgentCosTest.html in any modern browser — no server, no dependencies, no login required.

This tool produces planning estimates only. Results are not a billing commitment. Actual charges depend on runtime behavior, orchestration paths, tenant configuration, model usage, Microsoft licensing changes, and feature availability.

What it covers

Agent type Billing model
Custom agent (Copilot Studio) Copilot Credits — per classic answer, generative answer, agent action, graph grounding, AI prompt, agent flow, content processing
Agent Builder agent (M365 Copilot declarative) Copilot Credits — knowledge sources only; public website grounding is free
SharePoint agent Copilot Credits — SharePoint via tenant graph grounding
Foundry agent (Azure AI Foundry Agent Service) Token-based — input + output tokens billed to Azure subscription; no Copilot Credits

How to use

  1. Open AgentCosTest.html in a browser.
  2. Select your agent type at the top.
  3. (Optional) Pick a quick-start template to pre-fill a realistic enterprise scenario.
  4. Fill in the steps:
    • Knowledge sources — which sources your agent uses and how many.
    • Components (custom agents) — topics, tools, AI prompts, flows.
    • Conversation profile — how many turns per conversation and what happens in each.
    • Test plan scale — number of scenarios × iterations.
    • Foundry agents — model, token sizes, turns, and built-in tools (File Search, Code Interpreter).
  5. Review the Example Prompt & Cost Trace to validate the per-turn breakdown.
  6. Review the Production Cost Estimate panel for totals, a credit breakdown chart, and capacity impact.
  7. Use Export CSV to save the configuration and results, or Print for a shareable snapshot.
  8. Click Start Over (header or results panel) to reset everything.

Quick-start templates

Template Agent type What it models
Enterprise FAQ agent Custom (Copilot Studio) SharePoint + enterprise connectors, authored topics, no tools or flows
HR policy agent Custom (Copilot Studio) SharePoint, connectors, uploaded files, pure internal knowledge — no tools, flows, or AI prompts
IT helpdesk with tools + flows Custom (Copilot Studio) SharePoint, connectors, tools, agent flow, AI prompts, reasoning model
Employee Self-Service — HR agent starter Custom (Copilot Studio) SharePoint HR policies + ServiceNow Knowledge KB + Workday connector; HRSD topics for create/get/update HR cases; shared orchestrator flow; no reasoning; 70% M365 user share
Employee Self-Service — IT agent starter Custom (Copilot Studio) SharePoint IT docs + ServiceNow Knowledge KB + Microsoft Self-Help connector; ITSM topics for create/get/update tickets; reasoning model for diagnostic classification; 50% M365 user share
General purpose assistant (GPT-4o) Foundry agent GPT-4o, 5 turns with file search, token-based billing
Code assistant (GPT-4.1) Foundry agent GPT-4.1, 6 turns with code interpreter, token-based billing

Billing rates reference

Copilot Studio / M365 Copilot agents

Source: Copilot Credits billing rates (Microsoft Learn)

Feature Credits Unit
Classic answer 1 per response
Generative answer 2 per response
Agent action (tool call, flow trigger) 5 per action
Tenant graph grounding 10 per message
SharePoint or connector (graph-grounded) 12 per query (10 graph + 2 gen)
Agent flow actions 13 per 100 actions
AI tools — basic 1 per 10 responses
AI tools — standard 15 per 10 responses
AI tools — premium 100 per 10 responses
Content processing 8 per page
Reasoning model surcharge +10 per generative answer and agent action

Pricing options (all bill in Copilot Credits):

Option Effective $/credit Notes
Pay-as-you-go (PAYG) meter $0.0100 Postpaid; billed via Azure subscription. No commitment.
Copilot Credit pack (subscription) $0.0080 $200/month for 25,000 credits. Unused credits do not roll over.
Copilot Credit Pre-Purchase Plan (P3) $0.0095 → $0.0080 1-year commit, 9 tiers, 5% → 20% discount (3,000 → 3,000,000 CCCUs). 1 CCCU = 100 credits.
Microsoft Agent Pre-Purchase Plan (P3) (NEW May 2026) $0.0095 / $0.0090 / $0.0085 1-year commit, 3 tiers, 5% / 10% / 15%. Covers BOTH Copilot Studio AND Microsoft Foundry usage. 1 ACU = 100 credits or $1 Foundry.

Source: Microsoft Copilot Studio Licensing Guide (May 2026).

M365 Copilot licensed users consume zero credits for all features when operating under their authenticated M365 Copilot identity (subject to fair-use limits).

Voice agents (introduced in the May 2026 Licensing Guide) are billed by total call length to the nearest second plus the configured voice orchestration. Voice agents are not modeled in this tool — estimate them separately.

Agent Builder and SharePoint agents using pay-as-you-go: billing is configured in the Microsoft 365 admin center and billed to the linked Azure subscription under Microsoft 365 Copilot pay-as-you-go — not Copilot Studio meters.

Azure Foundry agents

Source: Azure OpenAI pricing · Foundry Agent Service overview

Billed by tokens (input + output) to your Azure subscription. Context accumulates each turn — every user message re-sends the full conversation history as input, so costs grow with conversation length.

Selected models (Global Standard pay-as-you-go, USD / 1M tokens):

Model Input Output
GPT-4.1-nano $0.10 $0.40
GPT-4o mini $0.15 $0.60
GPT-5-nano $0.05 $0.40
GPT-5-mini $0.25 $2.00
GPT-5.1-codex-mini $0.25 $2.00
GPT-4.1-mini $0.40 $1.60
GPT-4.1 $2.00 $8.00
GPT-4o (2024-11-20) $2.50 $10.00
o4-mini $1.10 $4.40
o3 $2.00 $8.00
GPT-5 (2025-08-07) $1.25 $10.00
GPT-5.1 $1.25 $10.00
GPT-5.2 $1.75 $14.00
GPT-5.3 $1.75 $14.00
GPT-5.4-nano $0.20 $1.25
GPT-5.4-mini $0.75 $4.50
GPT-5.4 (<272k ctx) $2.50 $15.00
GPT-5.4 Pro (<272k ctx) $30.00 $180.00

Built-in tools: File Search $2.50/1K calls · Code Interpreter $0.033/session.

Other resources

1 Copilot Credit = $0.01 USD

The Microsoft agent usage estimator forecasts monthly production usage. This script does the opposite — it calculates the cost of your test plan based on individual conversations, so you can budget before you start testing.

Prerequisites

  • PowerShell 7+ (the script uses #Requires -Version 7.0)
    • Check your version: $PSVersionTable.PSVersion
    • If you need to install it: winget install Microsoft.PowerShell
  • No modules or authentication required — everything runs locally

Quick Start

Option A: Interactive Mode (guided prompts)

.\Estimate-CopilotStudioTestCost.ps1

The script will walk you through each test scenario step by step.

Option B: Batch Mode (from a CSV file)

.\Estimate-CopilotStudioTestCost.ps1 -ScenarioFile .\sample-test-plan.csv -PrepaidCredits 25000

Step-by-Step Guide: Interactive Mode

Step 1 — Open a terminal and navigate to the script

cd path\to\copilot-studio-test-cost

Step 2 — Run the script

.\Estimate-CopilotStudioTestCost.ps1

Step 3 — Enter your prepaid capacity (optional)

The script asks for your monthly prepaid Copilot Credits. This lets it show what percentage of your capacity testing will consume. Enter 0 to skip.

Monthly prepaid Copilot Credits - enter 0 to skip capacity check [0]: 25000

Step 4 — Define your first test scenario

Give the scenario a descriptive name:

Scenario name: FAQ flow test

Step 5 — Enter the features used in ONE test conversation

For each feature, enter how many times it fires in a single conversation. Press Enter to accept the default of 0.

Classic answers [0]: 2
Generative answers [0]: 3
Agent actions - triggers / topic transitions / deep reasoning [0]: 1
Tenant graph grounded messages [0]: 0
Agent flow actions - note: FREE in test pane [0]: 0
AI prompt tool calls - basic tier [0]: 0
AI prompt tool calls - standard tier [0]: 0
AI prompt tool calls - premium tier [0]: 0
Content processing pages [0]: 0
Uses reasoning model? [y/N]: n

Tip: If you're unsure what counts as each feature, open your agent in Copilot Studio, run a test conversation, and check the activity map — each step shown there maps to the features above.

Step 6 — Enter the number of iterations

How many times will this scenario be run? Count all testers and all passes (e.g., 3 testers × 2 passes = 6).

How many times will you run this scenario? [1]: 10

Step 7 — Add more scenarios or finish

Add another scenario? [y/N]: y

Repeat steps 4–6 for each scenario. When done, enter n.

Step 8 — Review the results

The script outputs a summary with:

  • Per-scenario breakdown — credits per conversation × iterations
  • Feature breakdown — which features drive the most cost
  • Capacity impact — percentage of your monthly prepaid credits consumed by testing

Step-by-Step Guide: Batch Mode (CSV)

Step 1 — Create your test plan CSV

Copy sample-test-plan.csv and edit it with your scenarios. The columns are:

Column What it means Example
ScenarioName Description of the test case FAQ flow test
ClassicAnswers Static/authored responses per conversation 2
GenerativeAnswers AI-generated responses per conversation 3
AgentActions Triggers, topic transitions, deep reasoning steps 1
TenantGraphMessages Messages using tenant graph grounding 0
AgentFlowActions Flow actions (FREE in test pane — tracked for awareness) 0
BasicPrompts Basic-tier AI prompt tool calls 0
StandardPrompts Standard-tier AI prompt tool calls 0
PremiumPrompts Premium-tier AI prompt tool calls 0
ContentPages Pages processed by content processing tools 0
UsesReasoningModel 1 if reasoning model is enabled, 0 otherwise 0
Iterations Total times this scenario will be run (testers × passes) 10

Example row:

FAQ flow test,2,3,1,0,0,0,0,0,0,0,10

Step 2 — Run the script with your CSV

.\Estimate-CopilotStudioTestCost.ps1 -ScenarioFile .\my-test-plan.csv

Step 3 — Add prepaid capacity (optional)

To see the capacity impact:

.\Estimate-CopilotStudioTestCost.ps1 -ScenarioFile .\my-test-plan.csv -PrepaidCredits 25000

Understanding the Output

===============================================================
  COPILOT STUDIO - TEST COST ESTIMATE
===============================================================

  Scenario Breakdown:
  -------------------------------------------------------------
  Scenario                       Per Conv  Iter.    Credits        USD
  -------------------------------------------------------------
  Greeting and FAQ                      9      5         45      $0.45
  Knowledge retrieval (ShareP...       16     10        160      $1.60
  ...
  TOTAL                                                2072     $20.72
Column Meaning
Per Conv Credits consumed by one execution of the scenario
Iter. Total number of times the scenario runs
Credits Per Conv × Iter.
USD Credits × $0.01

The Credit Breakdown by Feature section shows which features cost the most — useful for identifying where to optimize.

The Capacity Impact section (when you provide -PrepaidCredits) shows:

  • What percentage of your monthly allocation testing will use
  • How many credits remain for production
  • The overage enforcement threshold (125% of capacity)

How to Count Features in Your Agent

Not sure what numbers to enter? Here's how to figure it out:

  1. Open your agent in Copilot Studio
  2. Run a test conversation in the test pane
  3. Check the activity map — each step maps to a feature:
What you see in the activity map Feature to count
A static, authored response Classic answer
An AI-generated answer from knowledge Generative answer
A trigger firing, topic transition, or deep reasoning step Agent action
A response grounded in Microsoft Graph tenant data Tenant graph grounding
A flow executing a sequence of steps Agent flow action (free in test pane)
A prompt tool running (check the tier in Prompt Builder) Basic / Standard / Premium prompt
Document/image processing Content processing (per page)
  1. Check if reasoning is enabled — in your agent's model settings, if a reasoning-capable model is selected, set UsesReasoningModel to 1

Billing Rates Reference

Source: Copilot Credits billing rates

Feature Credits Unit Free in Test Pane?
Classic answer 1 per response No
Generative answer 2 per response No
Agent action 5 per action No
Tenant graph grounding 10 per message No
Agent flow actions 13 per 100 actions Yes
AI tools (basic) 1 per 10 responses No
AI tools (standard) 15 per 10 responses No
AI tools (premium) 100 per 10 responses No
Content processing 8 per page No

Reasoning model surcharge: When enabled, adds 10 credits per generative answer response (100 credits / 10 responses) on top of the base rate.

Tips to Reduce Test Costs

  • Start without tenant graph grounding — at 10 credits/message, it's the most expensive per-unit feature. Test your agent's core logic first, then enable graph grounding for a final validation pass.
  • Test with standard models first — if your agent uses a reasoning model, switch to standard for initial testing (2 credits vs 12 credits per generative answer). Use reasoning only for final QA.
  • Prioritize critical paths — don't iterate every scenario equally. Run edge cases 2–3 times, but run your happy path and critical flows more.
  • Leverage free flow testing — agent flow actions don't consume credits in the test pane, so test flows thoroughly before publishing.
  • Combine similar scenarios — if two test cases exercise the same features, merge them to reduce overhead.