Costs and budgets

Each step records tokens in and out and its cost. Amounts are tracked as integer micro-USD, so there are no rounding surprises.

Price table

Prices come from a configurable table:

[
  { "provider": "openai", "model": "gpt-4.1-mini", "inputPerMTok": 0.4, "outputPerMTok": 1.6 },
  { "provider": "bedrock", "model": "anthropic.*", "inputPerMTok": 3, "outputPerMTok": 15 },
  { "provider": "mcp", "model": "*", "inputPerMTok": 0, "outputPerMTok": 0, "perToolCallUsd": 0.001 }
]

The values above are examples; use your contract prices. Exact model names win over globs. Prices are looked up by provider name first, then by provider kind. simulated/* and ollama/* cost nothing. A call without a matching price is recorded with priced: false, so gaps are visible.

Budgets

budget:
  maxTokens: 50000
  maxCostUsd: 0.5
  maxSteps: 12
  maxToolCalls: 6
  timeoutSeconds: 300

Budgets can be set for the pipeline and for each agent; the stricter value wins. When a budget is used up, the control agent stops the run.

Coming with the API Monthly budgets per agent and team with alerts, and a cost dashboard in the console.

Attribution and export

Roadmap Every cost line carries tenant, agent, use case, run, step, model and provider. Totals can be aggregated by any of these and exported as CSV or JSON; a Prometheus metric with bounded labels covers dashboards. Hard-stop budgets can be set per run, agent, use case, tenant and month. Prices come from the pinned catalog described in your keys, your models.