Skip to main content
Apragya AIFree quotation builderQuotation builderQuote builderno signup
Sign inRegister
AI Credits

AI Credits and Usage Metering: One Unit for Every AI Feature, Visible by User, App and Company

For finance heads and admins who want AI used widely, with a clear limit, a clear owner for every unit spent, and no surprise bill.

Start freeTalk to sales
A finance admin adjusting taps that share water from one tank

AI credits and usage metering answer a question finance teams now ask every month: who used how much AI, on what, and is it worth it? Model providers bill in tokens, and each model prices them differently.

When a business uses chat, document reading, agents and app building across several models, those token bills are hard to read and harder to assign to a team.

Apragya meters every AI feature in one unit, the AI Credit. This page explains how it works, with an illustrative example.

What unmetered AI use costs

Surprise bills

A heavy week of document reading shows up only on the invoice.

No owner

Usage cannot be traced to a person, an app or a company.

Apples and oranges

Different models price tokens differently, so totals mislead.

Late warning

Nobody knows the budget is nearly spent until work stops.

LLM cost control that finance can read

Setting a monthly spend cap with a provider stops the bill, but it also stops work without warning, and it says nothing about who used what. A metered unit with allocations and alerts lets you keep AI running where it pays for itself.

Provider bills against Apragya AI Credits

FeatureRaw provider billingApragya AI Credits
UnitTokens, priced per modelOne credit, weighted by model
Check before use✗ NoBalance checked before each call
Per-user limitsRareAllocations per user and app
BreakdownBy API keyBy user, app, model, agent and company
AlertsSpend capsPlan percentage and unusual burn

Checked against public information on 16 September 2026.

An example: an Ahmedabad chemicals trader

Take an illustrative trader with 22 users on AI tools. In one month, the purchase team began reading every supplier certificate with AI, and usage jumped. The admin needed to see why before the pool ran out.

Where one month's AI Credits went

Organisation pool

The credits included with the plan.

  • By user

    Two purchase executives used the most.

  • By app

    Document reading led the month.

  • By agent

    Each agent's runs and their usage.

  • By model

    Heavier models weigh more per token.

  • By company

    Stamped on every usage row.

  • Alerts

    A warning at the plan percentage set.

The breakdown showed the jump came from one useful new habit, not waste.

The admin saw the spike came from certificate reading, which saved hours of typing. She raised the purchase team's allocation, trimmed two idle allocations, and set an alert well before the pool would run out.

Bowls of tokens around a central jar, one fuller than the others
One bowl was fuller for a good reason, so its allocation went up.

How AI usage tracking works

Step 1 of 5

Weighted tokens

One AI Credit equals 1,000 weighted tokens; lighter models weigh less.

  1. Weighted tokens: One AI Credit equals 1,000 weighted tokens; lighter models weigh less.
  2. Checked first: Before a call, the organisation pool and the user's allocation are checked.
  3. Allocated: Admins give each user an allocation per app; no allocation means no usage.
  4. Recorded: Each call is logged with user, app, model and company.
  5. Alerted: Alerts fire on plan percentage or unusual burn, by email, Slack, Teams or webhook.

What sets AI Credits in India apart

Plan-based pool

Credits come with the plan, so spend is known in advance.

Per company

Multi-company groups see usage by company.

Agents included

Agent runs and workflow steps are metered in the same unit.

Deny by default

A user without an allocation cannot spend credits.

Trends

Daily usage and a forecast for the period.

Your own keys

Organisations can add their own model provider keys.

For a department view, see tracking AI cost per department. For keys, see bring your own LLM key. Most usage comes from tools like the AI document summariser.

Getting started with AI Credits

  1. List the teams and apps that will use AI first.
  2. Give each user an allocation for the apps they need.
  3. Set an alert at a comfortable share of the pool.
  4. Review the breakdown by user and app after two weeks.
  5. Move allocation from idle users to busy ones.

Metering works best as a monthly conversation with each team, not as a hard stop.

What it costs to start

Free
Up to 3 users
CRM, HR and Vendor Onboarding
INR 3,999
Starter, a month
Up to 10 users, CRM and HR plus a choice of apps
INR 7,999
Pro, a month
Up to 25 users and every enterprise app

AI credits and usage questions

Why not just count tokens?
A token on a large model costs far more than on a small one. Weighting turns both into one comparable unit.
What happens when a user's allocation runs out?
Further AI calls for that user are refused until an admin raises the allocation or the period resets.
Are agent runs counted?
Yes. Agent runs and workflow steps are metered in AI Credits, like chat and document reading.
Can we see usage by company?
Yes. Each usage row carries the company that was active when the call was made.
What does it cost to start?
The free plan covers up to 3 users with CRM, HR and Vendor Onboarding and 1,000 AI Credits a month. Starter is INR 3,999 a month for up to 10 users, with CRM and HR plus a choice of apps such as Sales and POS, Inventory, Payroll, Purchase and Payables, Support and Manufacturing. Pro is INR 7,999 a month for up to 25 users with every enterprise app, including Finance and Accounting, plus 5,000 AI Credits, single sign-on and audit logs.

See where your AI Credits go

Give each user an allocation, set an alert, and read the breakdown after two weeks.

Start freeTalk to sales