Skip to content

Credits & Billing

Credits Billing

Credits are the unit of currency consumed when agents run on the nara platform. Each agent execution consumes credits proportional to the LLM tokens used during the conversation — including input tokens (the prompt and context), output tokens (the agent’s responses), and any tool call overhead.

Credits provide a predictable way to track and budget AI usage across your organization, regardless of which LLM model or agent is in use.

Navigate to Settings > Credits to view your organization’s current credit balance. The balance page shows:

FieldDescription
Available creditsThe number of credits currently available for use
Reserved creditsCredits that are currently reserved for in-progress agent runs
Total consumedThe cumulative number of credits consumed since the account was created

The credit ledger is a detailed, chronological history of every credit transaction in your organization. Each entry records:

ColumnDescription
TimestampWhen the transaction occurred
TypeThe transaction type: consumption, reservation, release, correction, or purchase
AmountThe number of credits (negative for consumption, positive for additions)
Balance afterThe running balance after this transaction
AgentThe agent associated with this transaction (if applicable)
SessionThe session or run that triggered the transaction
DescriptionA brief explanation of the transaction

Use the ledger to audit credit usage, identify unexpected consumption patterns, and verify billing accuracy.

Monitor how credits are consumed across different dimensions:

See which agents consume the most credits. This view ranks agents by total credit usage over the selected time period, helping you identify high-cost agents that may benefit from prompt optimization or model selection changes.

The daily usage view provides a breakdown of credit consumption for each day:

  • Total credits consumed — aggregate consumption for the day.
  • Number of agent runs — how many agent executions occurred.
  • Average credits per run — the mean cost of each agent execution.
  • Peak hour — the hour with the highest credit consumption.

Usage distribution charts visualize these statistics as graphs and heatmaps, making it easy to spot trends and anomalies at a glance.

Credits follow a reservation-based flow to ensure accurate accounting during agent runs:

  1. Reserve — when an agent run begins, credits are reserved from your available balance based on an estimated maximum consumption. This prevents over-spending if multiple runs execute simultaneously.

  2. Agent runs — the agent processes the conversation, invoking tools and generating responses. Actual token usage is tracked throughout the run.

  3. Commit — when the run completes, the actual credit cost is calculated based on real token usage. The committed amount (which is typically less than the reservation) is permanently deducted from your balance.

  4. Release — any unused portion of the reservation is released back to your available balance. The difference between the reserved amount and the committed amount is returned.

Available Balance: 1000
↓ Reserve 100 for agent run
Available: 900 | Reserved: 100
↓ Agent run completes (actual cost: 42 credits)
↓ Commit 42, release 58
Available: 958 | Reserved: 0

In rare cases, an administrator may need to correct the credit balance manually — for example, to compensate for a platform error or to apply a promotional credit. Corrections are recorded in the ledger with a clear description and are visible in the transaction history for auditing.

When your credit balance runs low, you can purchase additional credits:

  1. Navigate to Settings > Credits.

  2. Click Purchase Credits.

  3. Select a credit package or enter a custom amount.

  4. Complete the purchase through the billing flow.

  5. The purchased credits are added to your available balance immediately and appear as a transaction in the ledger.