Skip to content

AI-agent costs

Log what AI agents cost, add your markup, and bill it next to the hours.

Get agent runs into eBRBe

The agent-cost ledger (every plan, including Free) records what Claude Code, Cursor, Codex and your own agents cost per run, in tokens and dollars, and totals it per client. Runs arrive in several ways:

  • Hooks and the ebrbe CLI for Claude Code and Codex, which report each session's usage.
  • OpenTelemetry spans from your own agents (see Developers).
  • Usage sync from the Anthropic and OpenAI admin APIs, daily, with your own admin keys (Pro and above).
  • Import a usage CSV from Claude Code, Cursor, the provider consoles or your own logs.
  • The REST API and the MCP server.

Costs are worked out from the model price list (the same one the free AI agent cost calculator uses).

Spend per client

Agent costs → Runs shows the totals for the period and a By client table: each client's agent cost (directly, or through the project's client) and how much of it is billable and not on an invoice yet. Amounts are at cost, before markup.

Markup

Markup rules and agent cost reports are on Pro and above. Agent costs are billed at cost unless you add a rule. Markup rules on a client or project charge, say, cost + 30%. The Agent costs page shows cost and the billable amount after markup per project.

Billing agent costs

When you create an invoice, billable agent runs for that client and period are added as lines next to the hours. Agent costs are in US dollars. Our guide on how to bill clients for AI agent costs covers pass-through, blended pricing and markup.

Who sees runs

You see your own runs, managers also see their projects' runs, and owners and admins see all of them.