Platform · platform.routerlab.buzz

The same bay your agents already call.

Usage, spend, requests, TTFT, and token counts at key and team scope. Aliases, waterfalls, and budgets are config — not a deploy.

Observe · August demo LIVE
BY AGENT · NOT A LIVE TENANT support-agent $1,904 research $1,212 triage $495

Keys

One key per caller. People, agents, and tools each get a scoped gateway key. Provider secrets never leave the rack.

Aliases

Call prod-chat in code. Route it to a hosted model today, your fine-tune tomorrow. Dev and prod are separate names.

Scopes

Per person, per agent, per tool. A scope can pin a caller to local-only models so sensitive work never leaves your GPUs.

Limits

Caps by day, week, or month — key, project, or team. A capped key pauses and says so.

Aliases on the board

Alias Routes to
prod-chatsonnet-class · hosted → backup
cheap-triagegemini-flash · google
draftshaiku-class · hosted
local-embedqwen-27b · your gpus local

Usage, two cuts

The same spend, two cuts: which agent burned it and which person owns it. p50 TTFT sits next to cost, so a slow key is as visible as an expensive one. Everything the dashboard shows is queryable over the Usage API.

Get API key Read the docs