Terminology

Millwork uses one word for each part of a run across the docs, the API examples, and the dashboard. This page defines each word, names the API field where it differs, and says what the word does not mean. A run is one task you send to Millwork; the API calls it an execution. A live run is submitted to call a model. Every live run has a platform fee, charged when Millwork accepts the run; read Model access and billing for the full rule.

What you save and what Millwork picks

Term	Meaning	What it is not
Arm	A saved option Millwork may run your task on: a model, an agent endpoint, or a skill. The docs say saved model; the API calls every saved option an `arm`.	Saving an arm does not force Millwork to pick it. An arm is not the route chosen for a run.
Catalog model	A model your organization can save now, listed by `GET /v1/model-catalog`.	A catalog entry is not a promise that the provider has capacity at this moment.
Route	The path recorded for a run: which saved model was chosen and any fallback options.	A route is not a second name for an arm. A test run makes no model call. Provider-reported model details are separate and may be absent.
Output check	A check applied to the run’s output. The API calls it a verifier (`verifier_id`). If you do not name one, a basic check confirms only that output is present.	An output check does not decide whether the answer is right for your product.
Provider account	Your account with the company, cloud, or gateway that serves a model. Its connection to Millwork decides who bills the model usage.	Millwork never returns or reveals your provider key.

A run and its records

Term	Meaning	What it is not
Run	One task you send to Millwork and follow to a final status. The API calls it an `execution`, with an `execution_id`.	A run is not proof that a model answered. A test run is also a run.
Test run	A run that exercises your integration without calling a model. The API calls it Echo (`mode: "echo"`). It has no platform fee and no model cost, and it counts toward your quota like any run.	A test run is not proof that a model can answer your task. It returns no model output.
Request key	The value you send in the `Idempotency-Key` header. A retry with the same key and the same body returns the same run instead of starting a second one.	A request key is not a run id. A retry with the same key and a different body is refused; read Errors and retries.
Result	The model output of a completed live run, from `GET /v1/executions/{executionId}/result`. It can expire.	Test, failed, cancelled, and expired runs have no result.
Receipt	The record of a run’s outcome, route, model usage, checks, and any follow-up action. Runs on catalog models also record model and provider details when available.	A receipt never contains the prompt or the output. Its model-usage totals exclude the platform fee.

Money and accounts

Term	Meaning	What it is not
Platform fee	Millwork’s charge for a live run, charged when Millwork accepts the run.	The platform fee is not model usage. Cancelling a run does not return it. Millwork returns it only when it fails a run before any model attempt starts.
Organization	Your company or team in Millwork: its members, API keys, saved models, and runs. The CLI and API call an organization a tenant.	An organization is not a person. API keys belong to the organization, not to a member.

Use How Solver works for the full sequence. Use Results and receipts for storage and expiry guidance.

Start here

Understand Solver

Principles and scope

What you save and what Millwork picks

A run and its records

Money and accounts

​What you save and what Millwork picks

​A run and its records

​Money and accounts

What you save and what Millwork picks

A run and its records

Money and accounts