Skip to main content

AI spend

See your AI bill. Cap it. Move on.

TokMeter shows AI spend across OpenAI, Anthropic, Azure, Vertex, Bedrock and more, plus tools such as Cursor, GitHub Copilot, Claude Code and n8n. Setup starts with a read-only key.

Built for engineering leaders. No code change to start. Read-only setup.

Three steps. Skip any.

  1. See the bill. Paste a read-only billing key and see spend by team, model and project.
  2. Track per request. Add the SDK with one import to attribute each call to a user, agent or CI job.
  3. Cap the spend. Point your OpenAI base URL at TokMeter for hard dollar caps per team. Roll back with one environment variable.

What you get

  • Per-team rollup

    Every dollar mapped to a team, model and project, with per-engineer cost attribution.

  • Virtual keys with hard caps: a key per team, agent or CI job, with a monthly ceiling.

  • Full audit trail of key, budget and credential changes, exportable.

  • Production evals on sampled real traffic.

  • Agent and MCP tracking: cost and policy per tool, per call.

  • Privacy controls: PII redaction and configurable retention.

Read-only by default

Billing keys, never write scopes. Default trace retention is 30 days, configurable per organization.

Common questions

What does TokMeter need to start?

A read-only billing key. No code change to start.

Can TokMeter cap spend?

Yes. Point your OpenAI base URL at TokMeter for hard dollar caps per team. Roll back with one environment variable.

How long are traces kept?

Default trace retention is 30 days, configurable per organization.

See your AI spend before your next invoice.