You pay for
green, not for
tokens.

• Early access

  Free, while we build

  • The gm loop, end to end
  • Bring your own runtime
  • Direct line to the people writing it
  • Your specs shape which runtime lands next

• Individual

  Planned · metered by verified run

  • Verified runs, billed only when they go green
  • Full run records, retained and exportable
  • Every supported runtime
  • No seat count, no per-editor licence

• Team

  Planned · annual

  • Shared run history across a team
  • CI-shaped usage: exit codes, no interactive session
  • Private endpoint, isolated from other tenants
  • Volume commitments instead of metering

Why this unit

Per-token pricing pays a model to be verbose, and per-seat pricing pays it to be installed. Neither is correlated with whether you got working code. Because gm already knows, mechanically, whether a run passed, we can bill the thing you actually wanted. We absorb the cost of the rounds where the model was wrong.

That also lines the incentives up on our side of the table: every round we waste is a round we paid for. Getting to green in two rounds instead of five is our margin, not your invoice.

What is not in the plan

• No training on your code by default. Runs we learn from are runs you opted in on.
• No per-editor licence. It is a command; it runs where your code runs.
• No charge for a red run. If it never passed, it never billed.