> ## Documentation Index > Fetch the complete documentation index at: https://docs.hopscotchlabs.ai/llms.txt > Use this file to discover all available pages before exploring further. # Glossary > The words this site uses for Hopscotch's own concepts, defined once and used the same way everywhere. These terms carry money and permissions, so they are defined once here and used identically on every other page. ### API key A credential you use to call `https://api.hopscotchlabs.ai`. It belongs to a workspace rather than to a person, and it is never a provider credential. Shown once at creation, stored only as a one-way hash, and identified afterwards by its name, its prefix and its last four characters. See [Create an API key](/guides/create-an-api-key). ### Balance The credit a workspace has left right now. ### Capability One of five recorded answers about what a model can do, carried on every catalog entry: `streaming`, `tools`, `json_mode`, `vision` and `audio`. Each answers `yes`, `no` or `unknown`, and `unknown` means nobody has established the answer rather than that the answer is no. A capability describes a model; it does not gate a request. See [Model capabilities](/concepts/model-capabilities). ### Credit Prepaid spending power on a workspace, bought through checkout and reduced by usage at face value. Credit is bought above face value, because the platform fee is added at purchase. A purchase therefore has two figures: the amount charged to the card, and the credit granted. On this site, "credit" always means the second. ### Deprecation The date a provider stopped recommending a model. The model still answers. It is not the same as a sunset. See [Model lifecycle](/concepts/model-lifecycle). ### Invitation An emailed link that offers somebody a role in a workspace. Bound to one email address, valid for seven days, confers nothing until it is accepted, and may carry a key reservation. See [Invite and manage members](/guides/invite-and-manage-members). ### Key reservation A promised key attached to an invitation. It occupies one of the workspace's five key slots from the moment it is made, and becomes a real key only when its holder views it for the first time after accepting. A reservation that lapses or is cancelled was never a key, so ending it is not a revocation. ### Live model A model that is callable right now, because we hold working credentials for it and it has been enabled. Only live models are offered to you. See [Models](/concepts/models). ### Member A person in a workspace, holding exactly one role there. A person may be a member of more than one workspace. ### Model A single callable model, named by a model id. ### Model id The value you put in the `model` field of a request, written as a family and a model separated by a slash, for example `openai/gpt-4o-mini`. ### Owner The one member of a workspace who holds the money: the card, the buying of credit, and pausing requests. A workspace has exactly one owner, always. Ownership moves only by transfer, never by being granted, and the transfer demotes the outgoing owner in the same action. The owner is also the only member who can make somebody an admin. ### Pause A workspace-level stop on serving requests, set by the owner. Any member may read whether requests are paused; only the owner may change it. Pausing revokes nothing. ### Platform fee The percentage Hopscotch adds to the credit you buy, charged at the moment of purchase and never taken when a request is served. Currently 2.5%. The rate that applies to a purchase is fixed when checkout starts. See [Credits and billing](/concepts/credits-and-billing). ### Playground A dashboard screen that sends a real request to a real model, authenticated by your session and attributed to a key you choose. It spends real credit and stores no prompt or reply. See [Playground](/guides/playground). ### Provider A company whose models we call. We hold the account and the credentials with each one, and you never do. ### Reasoning tokens Tokens a model generates while working out its answer, before the answer itself. They are counted inside the completion count rather than added to it, they are priced as their own kind, and they are reported separately where the provider reports them at all. See [Reasoning models](/concepts/reasoning). ### Request id The identifier attached to every response, successes and errors alike. It is the one thing worth quoting when asking us about a specific request, and it is how a single request is looked up in your usage. ### Role What a member may do in one workspace: one of owner, admin, developer or viewer, and a member holds exactly one. A role is a set of permissions rather than a rank. See [Workspaces and members](/concepts/workspaces-and-members). ### Spend rate cap A ceiling on how much a workspace may commit to in-flight requests over a rolling window. It is the one limit you may raise yourself. See [Spend controls](/guides/spend-controls). ### Statement The line-by-line record of every change to a workspace's balance: what was bought, what was charged, and what each request took. Any member may read it; only the owner may act on money. ### Sunset The date a model stops answering. After it, the model leaves the catalog and a request naming it is refused. Distinct from a deprecation, which is advice rather than a shutdown. See [Model lifecycle](/concepts/model-lifecycle). ### Usage record The record of one request: when it was made, which model, which provider served it, the token counts, the cost, the latency, and the outcome. Never the prompt and never the reply. See [Usage and metering](/concepts/usage-and-metering). ### Workspace The unit you sign in to and invite people into. Named, and holding members, API keys, usage records, a playground, spend controls, and its own billing. Anything that spends belongs to a workspace.