Platform

Teams & budgets

Organizations, teams, members, budgets, and model allowlists.

TokenRouter’s control model mirrors how companies actually organize AI spend: an organization contains teams, teams contain members, and API keys belong to a team (and optionally a member).

Structure

• Organization — your account, subscription, and provider keys.
• Teams — a product, department, or environment (e.g. search-prod, data-science). Each team gets its own budgets, allowlist, and keys.
• Members — people invited into teams; usage can be attributed and budgeted per member.

Budgets

Budgets are dollar limits over a window — daily or monthly — and can attach to a team, a member, or an individual API key. Each budget is one of:

• Hard cap — once spend reaches the limit, further requests are rejected with 429 budget_exceeded until the window resets.
• Alert-only — requests keep flowing; owners get notified as thresholds pass.

Both kinds fire email alerts at 50%, 80%, and 100% of the limit, so a hard cap never lands as a surprise.

Model allowlists

Each team can restrict which models its keys may call. A request outside the allowlist fails with 403 model_not_allowed, and auto routing only ever selects allowed models. Typical use: pin production to a vetted set while a research team keeps the full catalog.

Per-key attribution

The rule of thumb is one API key per app. Because every request is tied to a key, and every key to a team and member, cost attribution falls out automatically — no tagging or client-side bookkeeping. Analytics and CSV exports break spend down by team, member, key, model, and provider.