← API reference

API key management

Use the management API to operate on API keys from your own code: list, create, update, revoke, and cap the API keys in your workspace — for example one spend-capped key per team member, created by your onboarding tooling. The same management key can also read your workspace's spend by day, key, end user and app (GET /api/v1/usage).

These routes authenticate with a management key, not a regular API key: Authorization: Bearer llmr_sk_.... Workspace admins create one at Settings → Management Keys (the secret is shown once). A management key is rejected on every model endpoint, so a leaked one can't spend your credits directly, and every call is scoped to the workspace that owns it; keys you create spend that workspace's credit balance.

{ "data": [ { "...": "key object" } ] }

curl https://videorouter.sh/api/v1/keys \
  -H "Authorization: Bearer $VIDEOROUTER_MANAGEMENT_KEY" \
  -H "Content-Type: application/json" \
  -d '{"name": "jane.doe", "limit": 20, "limit_reset": "daily"}'

{
  "key": "llmr_sk_live_...",   // shown once — store it now
  "hash": "9f2c4e1a...",
  "name": "jane.doe",
  "label": "jane.doe",
  "disabled": false,
  "limit": 20.0,
  "limit_reset": "daily",
  "limit_remaining": 15.7083,
  "usage": 4.2917,
  "usage_daily": 4.2917,
  "usage_weekly": 18.02,
  "usage_monthly": 61.5,
  "created_at": "2026-10-20T10:00:00+00:00",
  "last_used_at": "2026-10-29T08:14:09+00:00",
  "expires_at": null
}

{
  "hash": "9f2c4e1a...",
  "name": "jane.doe",
  "label": "jane.doe",
  "disabled": false,
  "limit": 20.0,
  "limit_reset": "daily",
  "limit_remaining": 15.7083,
  "usage": 4.2917,
  "usage_daily": 4.2917,
  "usage_weekly": 18.02,
  "usage_monthly": 61.5,
  "created_at": "2026-10-20T10:00:00+00:00",
  "last_used_at": "2026-10-29T08:14:09+00:00",
  "expires_at": null
}

{ "deleted": true }

curl "https://videorouter.sh/api/v1/usage?start=2026-10-01&end=2026-10-28&group_by=key,end_user" \
  -H "Authorization: Bearer $VIDEOROUTER_MANAGEMENT_KEY"

{
  "data": [
    { "date": "2026-10-28", "key_hash": "9f2c4e1a...", "key_name": "jane.doe", "end_user": "emp-1042",
      "requests": 38, "errors": 1, "cost": 4.2917, "input_tokens": 51200, "output_tokens": 9800 },
    { "...": "more rows" }
  ],
  "start": "2026-10-01", "end": "2026-10-28",
  "group_by": ["key", "end_user"], "granularity": "day", "timezone": "UTC",
  "total_cost": 112.4031
}

Spend limits

When a key reaches its limit for the current window, its requests return 402 until the window resets (or you raise the limit):

HTTP 402
{ "error": { "message": "daily spend cap reached ($20.00 of $20.00); resets 2026-10-30 00:00 UTC",
             "type": "insufficient_quota", "code": "spend_cap_exceeded" } }

• The same error on every billable endpoint — chat (streaming too), embeddings, images, video, audio, lip-sync, face swap, avatars, 3D, live video, and batches/jobs. The realtime WebSocket is refused at connect time and closes with code 4402.
• The cap is checked before a request starts, so the request that crosses the line still completes — a key can end a window over its limit by at most one request's cost. Chat requests near the cap get max_tokens clamped to what's left.
• In-flight video jobs count — the amount held for jobs still running counts against the cap immediately, so a key can't queue unlimited jobs while under the limit.
• A batch or job is checked against the remaining budget as a whole: it is refused if its estimated cost would push the key over the cap.

Tagging requests

Usage can only be split by end user and by app if requests carry the tags. Send them on every request; they are recorded on every model endpoint.

Fields

Advisory only — set by the client, never used for authentication or billing. For a hard per-person boundary, give each person their own key and a limit. See App attribution.

Errors

Status codes