Choose an estimator

Single agent AI Units estimator

Estimate the cost (in AI Units) for a single AI agent. Use this as a lower-end planning baseline; actual usage and cost may vary.

Session workload profile

Note: Values below represent usage per session. Document image counts can be updated.

A session is one user's continuous interaction with the single agent to complete a task or get answers to one or more questions.

Feature Value Reasoning steps Tokens/base AI Units

Cost estimates

This cost estimate does not subtract the pooled 20,000 AI Units monthly allowance. The allowance is shared across agents and apps and does not carry over.

Per session AI Units

0.00

Total daily AI Units

0

Total monthly AI Units

0

Based on 21.7 working days per month.

Estimated monthly cost

$0

1 AI Unit = $0.01

AI Unit packs can be purchased in multiples of 100,000 at $1,000 per pack.

The estimator does not include the 30 million pooled AI units provided with Agentic App SKU.

How to use

  1. Choose complexity Select the complexity that best matches the agent prompt, tools, and reasoning required for a typical question.
  2. Enter expected usage Update daily active users, sessions per user, and document image counts where applicable.

Assumptions

  • Feature values are defaulted based on the agent complexity.
  • Tokens consumed for each feature are estimates based on the agent complexity.
  • Same feature counts can use different AI Units when agent complexity changes, because more complex agents typically have richer prompts and more reasoning steps.

Disclaimer: The AI Units estimator is not a binding offer or a guarantee of the final cost or availability of the product. This estimate is only a guidance and not incorporated into a contractual agreement. The actual usage and associated cost may vary depending on the region, actual usage, number of users, your application configuration and other factors.

How to select complexity

  • How complex is your prompt?
  • How many records does a tool call return? More records indicate higher complexity.
  • How many tool calls are required for a typical user question? More than one usually means a high-complexity agent with more reasoning steps.
  • How complex is your summarization prompt?

Multi Agent and Workflow AI Units estimator

Estimate the cost (in AI Units) for multi-agent teams and workflows. Use this as a lower-end planning baseline; actual usage and cost may vary.

Session workload profile

Note: Values below represent usage per session. Worker agents and LLM nodes can be updated where applicable.

A session is one user's continuous interaction with a multi-agent team or workflow to complete a task or get answers to one or more questions.

Feature Value Reasoning steps Tokens/base AI Units

Cost estimates

This cost estimate does not subtract the pooled 20,000 AI Units monthly allowance. The allowance is shared across agents and apps and does not carry over.

Per session AI Units

0.00

Total daily AI Units

0

Total monthly AI Units

0

Based on 21.7 working days per month.

Estimated monthly cost

$0

1 AI Unit = $0.01

AI Unit packs can be purchased in multiples of 100,000 at $1,000 per pack.

The estimator does not include the 30 million pooled AI units provided with Agentic App SKU.

How to use

  1. Choose the agent type Select whether the estimate is for a supervisor-based multi-agent team or a workflow.
  2. Enter expected usage Update the "Value" field for number of worker agents and LLM nodes when editable.

Assumptions

  • Feature values are defaulted based on the agent complexity.
  • Tokens consumed for each feature are estimates based on the agent complexity.
  • Same feature counts can use different AI Units when agent complexity changes, because more complex agents typically have more worker agents or workflow nodes and richer prompts.

Disclaimer: The AI Units estimator is not a binding offer or a guarantee of the final cost or availability of the product. This estimate is only a guidance and not incorporated into a contractual agreement. The actual usage and associated cost may vary depending on the region, actual usage, number of users, your application configuration and other factors.

How to select complexity

  • How many worker agents do you have? More agents indicate higher complexity.
  • How many tools do you have across all agents? More tools indicate higher complexity.
  • How complex are the supervisor and worker agent prompts?
  • How many records do tool calls return? More records indicate higher complexity.
  • How many tool calls does a typical user question require? More calls indicate higher complexity.
  • How many agent calls does a typical user question require? More calls indicate higher complexity.

Agentic apps AI Units estimator

Estimate the cost (in AI Units) for your agentic app. Use this as a lower-end planning baseline; actual usage and cost may vary.

Consider workflow count, nodes per workflow, and prompt complexity. Use Medium if unsure.

Session workload profile

A session is one user's continuous interaction with the app to complete a task, including the initial load and follow-up actions.

Note: Feature counts below represent usage per app, per session.

Feature Value Tokens/base Action type AI Units

Cost estimates

This cost estimate does not subtract the pooled 20,000 AI Units monthly allowance. The allowance is shared across agents and apps and does not carry over.

Per session AI Units

0.00

Total daily AI Units

0

Total monthly AI Units

0

Based on 21.7 working days per month.

Estimated monthly cost

$0

1 AI Unit = $0.01

AI Unit packs can be purchased in multiples of 100,000 at $1,000 per pack.

The estimator does not include the 30 million pooled AI units provided with Agentic App SKU.

How to use

  1. Set baselines Use the top panel to define your app's complexity and projected user volume (DAU and sessions).
  2. Scale workloads Changing the "Number of agents" will automatically scale up your dependent initial load & user interaction features proportionally.
  3. Fine-tune data Update the "Value" fields to adjust the default counts.

Assumptions

  • App complexity sets the number of agents used by the app. You can change this if required.
  • Tokens consumed for each feature are estimates based on the app complexity.
  • Same feature counts can use different AI Units when app complexity changes, because more complex apps typically have more workflow nodes and richer prompts.
  • Initial load and user interaction feature values are either calculated from, or scaled by, the number of agents value.
  • The estimator does not include AI Units consumed by processes that run in the background.

Disclaimer: The AI Units estimator is not a binding offer or a guarantee of the final cost or availability of the product. This estimate is only a guidance and not incorporated into a contractual agreement. The actual usage and associated cost may vary depending on the region, actual usage, number of users, your application configuration and other factors.

How to select complexity

  • How many workflows are associated with the app?
  • How many nodes do you have in each workflow?
  • How many agent nodes do the workflows have?
  • How many LLM nodes do the workflows have?
  • How many child workflows are invoked from the main workflows?