Introducing TokenRouter: the financial ops gateway for LLMs

AI spend has a visibility problem. Tokens are bought by engineers, burned by agents, and reconciled by finance weeks later from invoices that say nothing about which team, feature, or developer the money went to. The tools that exist mostly observe the problem — dashboards, alerts, weekly reports — while the requests keep flowing.

TokenRouter is a gateway, which means it sits where the problem can actually be fixed: in the request path. Today we're opening it up with a 14-day free trial on every plan.

You point your existing OpenAI or Anthropic SDK at api.tokenrouter.io, bring your own provider keys, and get one tr_ key per app in front of 13 providers. Budgets with hard caps, per-key rate limits, and full cost attribution are enforced on every request — and usage analytics show requests, tokens, cost, and latency by team, member, key, model, and provider.

Compatibility is the part we're most opinionated about. OpenAI-shaped requests pass through to OpenAI verbatim; Anthropic requests hit a native /v1/messages endpoint, so Claude Code works by changing one environment variable. We don't rewrite your payloads unless you're crossing providers, and we never store your prompts — metadata only.

Flat plans at $29, $99, and $499 a month. No free tier, no usage component, no markup on tokens — your providers bill you directly at their rates. Every plan starts with a 14-day trial, card required, cancel anytime.

If you've been meaning to get AI spend under control before the next invoice surprise, this is the two-week window to try: swap a base URL, invite your teams, set a cap.