Configure LLM Pricing

Configure LLM Pricing

Use LLM Pricing when your contracted model rates differ from the default model pricing used by Mavvrik.

Pricing configuration changes the rate used for matching LLM usage. It does not instrument the application or create usage records.

Open LLM Pricing

Open the agent settings area and select Pricing.

Click + Pricing to create a pricing card.

Configure the pricing card

Configure:

• Pricing card name;

• Effective from;

• Currency;

• Apply to.

Apply to supports:

• Tenant;

• Workspace;

• Team(s);

• Agent(s).

When multiple published cards match the same LLM usage, Mavvrik applies the more specific scope:

Agent(s) > Team(s) > Workspace > Tenant

Conflicting published rules at the same specificity must be resolved before publishing.

Configure an LLM rate

Use the LLM Pricing tab for negotiated or customer-specific model rates.

Supported fields include:

• provider/platform;

• model name;

• optional region;

• pricing unit: per 1K tokens or per 1M tokens;

• input token price;

• output token price;

• optional cached input price;

• optional batch input/output prices.

For supported models, pricing can also include dimensions such as reasoning tokens, embeddings, image/audio/video usage, cache write, and long-context pricing.

Example: contracted model rate

Pricing card: OpenAI Contract Rates
Apply to: Tenant
Provider/platform: OpenAI
Model: gpt-4o-mini
Pricing unit: per 1M tokens
Input price: <contracted rate>
Output price: <contracted rate>
Effective from: <contract effective date>

Publish the card after the scope, model, pricing unit, and rates are verified.

Example: agent-specific rate

Apply to: Agent(s)
Agents: support-agent

The agent-specific rule takes precedence over matching Team, Workspace, or Tenant rules.

Pricing card states

• Draft — saved but not used for cost calculation.

• Published — eligible to price matching usage from the effective date.

• Disabled — no longer applied to future matching calculations.

Use Save as draft while preparing a card. Publish only after validating the scope, model, unit, and rates.

Verify an LLM pricing rule

After publishing:

1. Generate new LLM usage that matches the pricing rule.

2. Open Home → Agentic → Cost.

3. Inspect the matching model usage.

4. Confirm the expected scope and rate are applied.

5. If the cost is unexpected, verify the effective date, scope precedence, provider/platform, model, and pricing unit.

Troubleshoot LLM pricing