Overview

This API returns a cost indication - based on comparable deployments in the portfolio, the platform's usage index and the agent's own characteristics.

An indication, not an appraisal

The result is explicitly not a formal cost audit prepared under a certified FinOps standard. Every response therefore carries three pieces of information meant to be read together:

  • monthly_cost_usd - the point estimate
  • range_usd - the plausible corridor
  • confidence and basis - how reliable the figure is

A confidence: "low" based on 6 comparable deployments means: quote the range in the stakeholder conversation, not the point estimate.

Two ways to start

For an already-deployed agent, agent_id is enough. During an initial planning conversation, when no record exists yet, pass characteristics instead - agent type, model/provider, expected request volume and, if known, integration count and deployment tier. Setting both at once is not allowed.

When the data isn't enough

For rarely-used agent types, you'll get a 422 with code: "insufficient_data". That's an honest answer, not a malfunction. In that case, widen the time window via period_months up to 36 months, or fall back to the usage index.

Privacy for comparable deployments

Comparable deployments are only ever delivered anonymized: owning teams are rounded to the business-unit level, the rollout date to the month. That makes it impossible to trace back to an individual deployment.